Overcapacity has long been one of the most contentious issues in China’s trade relations with both the European Union and the United States. In Washington and Brussels, the prevailing view is that China has continuously directed large amounts of capital into strategic manufacturing sectors through subsidies, policy lending, low-cost land and energy, government procurement, and local industrial policies. Because domestic demand is seen as insufficient to absorb the resulting output, the surplus is ultimately exported at low prices, putting pressure on manufacturing industries in other countries.

That said, the United States and the European Union do not frame the issue in exactly the same way. The United States primarily views it through the lens of macroeconomic imbalances, industrial security, and strategic competition with China. The European Union, by contrast, focuses more on demonstrating the existence of specific subsidies, import injury, and the legal basis for trade remedy measures.

I have participated in several discussions on China-Europe economic relations, involving both think tanks and industry representatives, and one thing has become increasingly clear to me: there remains a significant gap between Chinese and European perspectives on this issue.

European experts generally argue that the challenge has evolved beyond isolated trade disputes or individual industries into a structural conflict between two different economic models. In their view, China relies on exports and industrial upgrading to sustain economic growth, while Europe sees its own manufacturing base coming under sustained pressure, with implications for employment, industrial capacity, and political stability. As a result, Europe’s “de-risking” agenda is driven as much by domestic economic and political considerations as by geopolitics.

Many also argue that years of dialogue have failed to change the underlying dynamics, prompting the EU to rely increasingly on legal and industrial policy tools in an effort to encourage adjustments by China. With China’s technological upgrading accelerating, many in Europe believe that waiting longer would only further erode Europe’s industrial competitiveness.

China apparently has a very different view, as indicated by the word “so-called” being used in the title of the official position paper. Chinese experts generally argue that the China-EU trade imbalance is primarily the result of macroeconomic structures rather than unfair competition driven by subsidies. China has maintained a high savings rate for decades, while domestic investment has slowed in recent years, naturally producing current account surpluses and capital outflows.

They contend that China’s manufacturing competitiveness is rooted in its enormous domestic market, highly integrated supply chains, and industrial clusters—not simply in government support. In this view, attributing trade imbalances entirely to industrial policy misdiagnoses the problem and risks turning what should be addressed through macroeconomic adjustment into a trade conflict.

Some Chinese experts also question the methodologies used in Western studies on China’s overcapacity and subsidy levels, arguing that they rely on strong assumptions and that policies built on those estimates could face challenges both economically and under international trade rules.

I am not a trade economist, and it will be too difficult to take a position on this debate. My impression is that this is an extraordinarily complex issue that deserves much more research and discussion by international experts.

The position paper released by the Chinese government today is, from my personal view, a detailed response to many of the criticisms that the United States and Europe have made against China on this issue over the past several years. I would recommend that anyone following China trade policy—including policy watchers and trade lawyers—read it carefully.

One small but telling detail is the title of the position paper. The official English version is "China's Position on the So-called Excess Capacity Issue," rather than using the much more common Western term "overcapacity." This is unlikely to be accidental.

In U.S. and European policy discussions, “overcapacity” has become a loaded political term, implicitly accepting the premise that China suffers from a structural overcapacity problem. By choosing the more neutral phrase “excess capacity” and prefacing it with “so-called”, China seems to signal that it does not accept the Western framing of the debate.

You can download the full English text of China’s position paper here:

China’s Position on the So-called Excess Capacity Issue

The full English transcript (my own translation) of today’s press briefing is available here:

Opening Remarks

Jia Huili, Deputy Director-General of the News Bureau of the State Council Information Office and Spokesperson:

Good afternoon, ladies and gentlemen. Welcome to this press conference of the State Council Information Office.

Today, the Ministry of Commerce released the document China’s Position on the So-Called “Overcapacity” Issue. We are joined by Mr. Yan Dong, Vice Minister of Commerce, who will brief us on the document and answer questions of concern. Also attending today’s press conference are Mr. Lin Weilong, Director-General of the Policy Research Department of the Ministry of Commerce; Mr. Han Yong, Director-General of the Department of World Trade Organization Affairs; and Mr. He Shaojun, head of the Department of Foreign Trade.

First, I will give the floor to Mr. Yan Dong.

Introductory Statement

Yan Dong, Vice Minister of Commerce:

Thank you, Madam Moderator. Friends from the media, good afternoon.

As just mentioned, the Ministry of Commerce today released the document China’s Position on the So-Called “Overcapacity” Issue. I would like to take this opportunity to briefly introduce the background to its release and its main contents.

In recent years, global economic growth has lacked momentum. The international community has paid broad attention to global supply-demand balance and capacity issues, and there has been much discussion. In particular, some economies, out of concern over their own industrial competitiveness and market position, have politicized economic and trade issues, hyped up the so-called issue of Chinese “overcapacity,” and even deliberately confused concepts by linking industrial subsidies, trade surpluses, economic imbalances, and market competition with “overcapacity.” They have pushed narratives such as “China Shock 2.0” and used the so-called “overcapacity” issue as a pretext to implement all kinds of protectionist measures.

President Xi Jinping has pointed out that blowing out others’ lamps will not make oneself brighter, and blocking others’ paths will ultimately block one’s own. China believes that capacity issues should be viewed in a comprehensive, objective, and fair manner, and must not be used as a pretext for protectionism. Doing so will only disrupt the global economic and trade order and the stability of industrial and supply chains, aggravate contradictions and divisions, and create long-term risks for global economic growth. To clarify the facts and explain China’s views and position on issues related to the so-called “overcapacity,” we prepared this position paper.

The Chinese version of the document contains more than 10,000 Chinese characters, and an English version has also been released simultaneously on the website of the Ministry of Commerce. The full text consists of three parts: a preface, the main body, and a conclusion. The main body has four chapters.

Chapter One is A Comprehensive and Objective View of Global Capacity and the So-Called “Overcapacity.” This chapter reviews, from a historical perspective, the evolution of the global capacity landscape and explains that the gradual shift of the global industrial pattern from a single center to multiple centers is the result of international industrial division of labor. “Overcapacity” is a dynamic phenomenon in a market economy. Different parties understand the concept differently, and the situation in different economies and industries should be measured according to their stages and levels of development.

Chapter Two is China’s Position on Four Relationships Related to “Overcapacity.” It objectively analyzes the relationship between industrial subsidies, trade surpluses, economic imbalances, market competition, and “overcapacity,” emphasizing that there is no necessary connection between industrial subsidies and overcapacity, that more exports and larger surpluses do not equal overcapacity, that global economic imbalance is a historical norm with complex roots, and that market competition is an important safeguard for the optimization and adjustment of capacity.

Chapter Three is China’s Commitment to Building a Modern Industrial System Through Openness and Cooperation. This chapter focuses on China’s policy practices and future direction. It explains that the rapid development of China’s modern industries is driven by innovation, and that the steady and healthy operation of industry depends on deepening reform. It also makes clear that the development of China’s modern industries is not the so-called “China Shock 2.0,” but rather “China Opportunity 2.0.”

Chapter Four is Jointly Promoting an Open and Inclusive Global Industrial and Supply Chain Cooperation Landscape. This chapter presents China’s vision and proposals. It calls on the international community to shoulder responsibilities together and move toward one another, advocate mutual benefit and win-win outcomes, respect market rules, strengthen coordination of industrial policies, expand market openness, create cooperation opportunities, uphold multilateralism, and build a fairer and more equitable international economic order.

That is all for my introduction. My colleagues and I are now ready to take your questions. Thank you.

Jia Huili:

Thank you, Mr. Yan Dong, for the introduction. We will now move to questions. Please identify your media outlet before asking your question.

Questions and Answers

Reporter from The Beijing News, Shell Finance:

At present, countries hold different views on the issue of “overcapacity,” and international institutions also lack a unified definition. Evaluation criteria vary significantly and there are major differences of opinion. How should we understand this issue? Thank you.

Yan Dong:

I will invite Director-General Lin Weilong of the Policy Research Department to answer this question.

Lin Weilong, Director-General of the Policy Research Department, Ministry of Commerce:

Thank you for your question. As you just mentioned, the concept of “overcapacity” and related issues have long been disputed in the international community, and no broad consensus has been reached. I would like to explain this from three dimensions so that the issue can be understood in a more comprehensive, objective, and systematic way.

The first is the theoretical dimension. “Overcapacity” is a complex concept that needs to be understood in a specific economic context. Most economists explain it from both macro and micro perspectives. At the macro level, it refers to supply exceeding demand, with total production capacity clearly surpassing total demand. At the micro level, it refers to idle enterprise capacity, with factors such as monopolistic competition preventing the clearing of capacity and keeping output below its optimal state. “Overcapacity” is a dynamic phenomenon in a market economy. It is related both to changes in supply and demand and to the industrial life cycle, and it always exists in a dynamic cycle of balance, imbalance, and rebalancing. Balance is relative; imbalance is common.

The second is the historical dimension. The global center of capacity has undergone several major shifts along with industrial relocation. Since the First Industrial Revolution, factors of production around the world have accelerated their movement, and the United Kingdom and the United States successively became world industrial centers. In 1880, the United Kingdom’s share of global industrial output reached a peak of 22.9 percent; in 1953, the United States’ share reached as high as 44.7 percent. After World War II, the world experienced multiple waves of industrial transfer: from the United States to Europe, from Europe and the United States to Japan, then to East Asia and China, and now from parts of China to Southeast Asia and other regions. This has formed three major regional manufacturing centers in North America, Europe, and East Asia. China’s emergence as the “world’s factory” is the result of actively integrating into economic globalization and participating in international industrial division of labor, and it is an important component of the global manufacturing network.

The third is the practical dimension. Capacity utilization needs to be assessed in light of the realities of different countries and industries. Internationally, capacity utilization is commonly used to measure “overcapacity.” This indicator is the ratio of actual output to potential output. Reasonable ranges differ across economies, and there is no universally applicable global standard. The median capacity utilization rate in developed and fast-growing economies is usually between 75 percent and 80 percent, while in less developed countries it is generally between 50 percent and 64 percent. At the same time, capacity utilization differs greatly across industries. In some traditional industries, capacity utilization is clearly below the average level. In some countries, utilization in industries such as beverages and furniture is around 65 percent, while in rubber, chemicals, and plastics it is only 40 to 50 percent. Emerging industries tend to have relatively higher rates. In some countries, capacity utilization in computers and peripheral equipment reaches 83 percent, and in electrical equipment and components it reaches 86 percent. Looking at countries’ economic practice, capacity utilization is somewhat objective as a measure of the state of capacity use, but it should not be absolutized as a standard for judging whether different economies or industries have overcapacity.

In light of China’s actual development, industrial capacity utilization as a whole remains within a reasonable range. In 2025, the capacity utilization rate of above-designated-size industry was 74.4 percent. In fields such as high-tech manufacturing, high-end equipment manufacturing, and strategic emerging industries, capacity was used more fully. In some traditional industries, such as raw materials, utilization was temporarily lower, mainly because of adaptive adjustments brought about by structural transformation and green transition. This is a normal phenomenon in the process of industrial upgrading and quality improvement. Overall, China’s industry remains basically balanced in supply and demand and operates steadily.

Thank you.

Reporter from China News Service:

In recent years, there has been an international narrative that subsidies cause overcapacity. How does China view such statements? Also, what is the relationship between industrial subsidies and “overcapacity”? Thank you.

Yan Dong:

Thank you for your question. I will ask Director-General Han Yong of the WTO Affairs Department to answer.

Han Yong, Director-General of the Department of World Trade Organization Affairs, Ministry of Commerce:

Thank you for the question. At present, many countries formulate industrial policies tailored to their national conditions and industrial development needs. For example, they provide research and development subsidies for emerging industries and risk subsidies for agriculture. These are WTO-consistent industrial and trade policy tools available to members. Multiple reports by UNCTAD show that the number of industrial policies worldwide has been on the rise over the past five years, and that R&D subsidies, tax incentives, and low-interest loans have become internationally common approaches to supporting emerging industries. What needs to be emphasized here is that industrial subsidies are not a problem in themselves, nor is there any necessary connection between industrial subsidies and overcapacity. Reasonable and compliant subsidy policies can help correct market failures, advance technological innovation, protect the ecological environment, reduce poverty, and promote balanced development. They do not create the so-called “overcapacity.”

On the other hand, if protectionist measures are adopted and non-compliant industrial policies are introduced to shift burdens onto others and restrict competition, they will disrupt the global economic and trade order. The U.S. Inflation Reduction Act plans to provide $750 billion in various subsidies from 2022 to 2031, and subsidized electric vehicles must meet conditions such as being produced and sold locally or in North America, thereby excluding other countries and regions. In addition, according to incomplete statistics, the European Commission will provide more than 1.44 trillion euros in various subsidies between 2021 and 2030. The EU’s Industrial Accelerator Act proposes to directly link local content with fiscal support through an “EU-origin” requirement, creating serious investment barriers and institutional discrimination. Major powers are the leaders of global industrial development and should set an example by using subsidies reasonably on the basis of openness, fairness, and compliance, avoiding discriminatory subsidy policies and avoiding artificial interference in the global layout of industrial and supply chains.

China has always strictly observed WTO rules and is committed to building and improving a fiscal subsidy system consistent with international practices. Over the years, China has taken multiple measures to continuously improve the compliance, scientific basis, and transparency of subsidies, and has constantly standardized and refined related policies. It has cleaned up and regulated some improper local subsidies and is studying the establishment of a unified negative list management mechanism for local fiscal subsidies. China fulfills its WTO transparency obligations in a timely and comprehensive manner, and its subsidy notifications cover the entire country. Chinese subsidies are mainly directed to areas such as scientific research and development, commercialization of technology, and market consumption. More often, China uses market-based and guiding means such as public services, technical standards, and skills training, with重点 support for technological innovation, small and medium-sized enterprises, green development, and energy conservation. China’s subsidies apply equally to all kinds of market entities. Foreign-invested enterprises actively participate and benefit on an equal footing.

Development is an eternal theme of human society. All countries should focus on making the global development pie bigger and introduce industrial policies in a reasonable and compliant manner, rather than using them as tools to restrain other countries’ development. China is willing to hold discussions with all parties under the WTO framework on relevant industrial policies, jointly regulate related practices, and promote the updating of multilateral rules.

Thank you.

Reporter from Bloomberg News:

Does China plan to address trade partners’ concerns about “overcapacity” and the displacement of domestic industries by low-cost Chinese exports? If so, what specific measures will it take? And do you expect the trade surplus to reach another record high this year? Thank you.

Yan Dong:

This question concerns trade, so I will ask Mr. He Shaojun from the Department of Foreign Trade to answer.

He Shaojun, head of the Department of Foreign Trade, Ministry of Commerce:

Thank you for your question. Trade surplus is a topic of broad concern, and I would like to take this opportunity to share some views from several angles.

First, a surplus reflects profound changes in the international division of labor. Looking back at the history of global economic development, countries generally went through stages of trade surplus during industrialization. Manufacturing powers such as the United Kingdom, the United States, Japan, and Germany all maintained trade surpluses for long periods. At the product level, 80 percent of U.S. chips are exported, and about two-thirds of commercial aircraft delivered by Boeing are sold to customers outside North America. In 2025, the European Union recorded surpluses of $92.2 billion in automobiles, $214.6 billion in pharmaceuticals, and $11.6 billion in cosmetics. More exports and larger surpluses do not equal overcapacity. China’s trade surplus reflects the completeness and efficiency of its industrial system and is an objective result of changes in the global division of labor and trade patterns.

Second, Chinese products meet the production and daily-life needs of countries around the world. Consumer goods produced in China, such as computers, mobile phones, furniture, clothing, and toys, provide consumers in various countries with more choices, lower consumption costs, and help buffer inflation risks. China’s exports of production equipment and intermediate goods strongly support the industrialization of its trade partners. For example, China supplies more than 80 percent of the world’s photovoltaic modules and 70 percent of wind power equipment, providing important support for the green transition of its trade partners. Foreign-invested enterprises contribute 16 percent of China’s surplus and have also earned substantial returns on their investment. This vividly shows that while the surplus is recorded in China, the benefits are shared by all parties.

Third, from the perspective of the overall balance of payments, although China runs a sizeable surplus in trade in goods, it has deficits in trade in services and in the capital and financial account. The current account surplus accounts for about 3.7 percent of GDP, which is within the internationally recognized reasonable range, and there is no significant external imbalance.

I would like to stress that China never deliberately pursues a trade surplus. We will continue to actively promote balanced development of imports and exports, continue opening our market to the world, and firmly expand imports. In the first half of this year, China’s import growth in goods trade was 22.1 percent, significantly faster than export growth. China remains the world’s second-largest import market, and its enormous market demand provides strong momentum for the development of its trading partners. Since last year in particular, we have built the “Export to China” brand and plan to hold more than 100 import promotion activities every year, with the aim of sharing China’s new development opportunities with the world. At many overseas “Export to China” events, Chinese enterprises have signed letters of intent with local companies, broadened sourcing channels for imports, and created more convenient pathways for companies from various countries to enter the Chinese market. These efforts have received positive responses and have generated mutually beneficial cooperation.

Going forward, we will take more practical measures to expand imports and share with countries around the world the new opportunities created by China’s large market.

First, we will continue to expand the influence of the “Export to China” brand. Through a series of matchmaking events, exhibition linkages, and targeted procurement, we will further increase imports.

Second, we will continue to strengthen policy support. We will improve the facilitation of import trade, optimize the sources and structure of imported goods, and better meet the needs of industrial development and people’s aspirations for a better life.

Third, we will make full use of various promotion platforms. We will continue to host major exhibitions such as the China International Import Expo and the China International Consumer Products Expo, regularly organize import-trade fairs for featured CIIE products, and better leverage national demonstration zones for import trade promotion and innovation, so as to actively drive the import of high-quality goods and services from around the world.

Thank you.

Reporter from International Market News, United States:

Recently, some leaders from EU countries and institutions have expressed concerns that China’s currency is undervalued and gives Chinese exporters an unfair advantage. What is the Ministry of Commerce’s response? Looking ahead to the second half of this year, especially amid continued trade tensions and the broader economic backdrop, how does the Ministry view the outlook for China-EU trade? Thank you.

Yan Dong:

Director-General Lin Weilong from the Policy Research Department will answer this question.

Lin Weilong:

Thank you for your question. We have taken note of the relevant developments. In recent years, under the strategic guidance of the leaders of China and the EU, China-EU economic and trade relations have developed generally well. In the first half of this year, trade in goods between China and the EU increased by 14.2 percent year on year, reaching $447.88 billion. The two sides have huge potential for cooperation in services trade, scientific and technological innovation, and the green economy. This is both the result of deep interconnection and complementarity in China-EU industrial and supply chains, and a reflection of the broader trend of global digital and green transformation and the future development of China and Europe.

At the same time, the EU has recently continued to roll out trade restriction measures against China and to hype the so-called issue of the renminbi exchange rate. This lacks a factual basis and is not conducive to the stable development of China-EU economic and trade relations. China adheres to giving the market a decisive role in exchange-rate formation and maintains the renminbi exchange rate basically stable at an adaptive and equilibrium level. As a responsible major country, China does not seek competitive advantage through currency depreciation. Chinese manufacturing is internationally competitive, and there is no need to promote exports through an undervalued exchange rate.

China is not the source of the EU’s economic and trade problems, but a partner in solving them. Protectionism leads nowhere; cooperation for mutual benefit is the right path. At the end of last month, Minister of Commerce Wang Wentao and European Commissioner for Trade and Economic Security Maros Sefcovic jointly chaired the first meeting of the China-EU trade and investment consultation mechanism, and reached consensus on the new positioning of the relationship as one of “stable and balanced key trade partners.” China is willing to use this mechanism to strengthen dialogue and consultation with the EU, properly handle differences and frictions, promote practical cooperation, advance trade toward a more balanced direction, and foster the steady and sound development of China-EU economic and trade relations, thereby injecting more certainty and positive energy into the global economy.

Thank you.

Reporter from National Business Daily:

There are currently two narratives in the international community regarding China’s industrial development and technological innovation: one is “China Shock 2.0” and the other is “China Opportunity 2.0.” How should we view these two narratives? Is China’s industrial development ultimately a shock or an opportunity for the world? Thank you.

Yan Dong:

Thank you for your question. I will answer this one myself.

This is indeed a topic that has attracted a great deal of attention and discussion. Some countries have put forward the so-called “China Shock 2.0” narrative, slandering China’s industrial development as a threat to the monopoly position of Western countries and as squeezing the development space of countries in the Global South. This narrative does not conform to the facts and cannot stand up to scrutiny.

For more than a decade, China has been an important engine of world economic growth, contributing about 30 percent. By leveraging its market strengths, industrial development, and technological progress, China has provided the world with growing “market dividends,” “development dividends,” and “innovation dividends.” These dividends overlap and reinforce one another, bringing the world more development opportunities and broader room for growth, which is what more rational and objective voices in the international community call “China Opportunity 2.0.” This can be understood from four aspects.

First, China’s industrial development is a ballast stone for the stability of global industrial and supply chains. China has the world’s largest industrial manufacturing system, with the most complete range of categories and the most integrated system. It continues to provide all kinds of industrial manufactured products in an efficient manner, strongly supporting stable global supply and effectively offsetting localized supply gaps caused by protectionism, geopolitical conflicts, and other factors. This has demonstrated China’s strong resilience and sense of responsibility, enabling it to play the role of a stabilizing anchor and major hub in global industrial cooperation. China’s exports of high-quality and cost-effective production equipment and components have lowered entry barriers for developing countries’ manufacturing sectors. From 2012 to 2024, China exported more than $30 billion in textile machinery to developing countries, helping some countries in Southeast Asia and South Asia become important textile producers and exporters.

Second, China’s industrial development is a new engine for global innovation cooperation. China adheres to innovation-driven development and has forged an effective path in which technological innovation leads industrial innovation and industrial upgrading in turn accelerates technological iteration. With the support of Chinese manufacturing, any valuable technological achievement can quickly be transformed into a real product, making China an ideal testing ground for both “0-to-1” validation and “1-to-N” scaling of new products and services. China remains open in innovation, and its rapidly growing innovative companies have delivered several-fold, and in some cases several-dozen-fold, returns to investors from around the world. Many of China’s innovations, including large AI models, follow an open-source path and are widely welcomed by countries everywhere. Global cumulative downloads of open-source large models have surpassed 10 billion, enabling more countries, especially developing countries, to access and afford new technologies.

Third, China’s industrial development is a driving force for the global green transition. China is accelerating a comprehensive green transition and promoting green, low-carbon industrial development. By the end of the 15th Five-Year Plan period, the scale of China’s green industries is expected to exceed 20 trillion yuan. The rapid development of China’s green industries and the growing supply of new energy products have enriched global supply and strongly advanced the global low-carbon transition. According to a report by the International Renewable Energy Agency, over the past decade, the global levelized cost of electricity for wind and solar projects has fallen by more than 60 percent and 80 percent, respectively, and a large part of that is attributable to Chinese manufacturing and Chinese capacity. At present, with global energy tensions and artificial intelligence driving up electricity demand, the International Energy Agency predicts that by 2030 global electricity consumption by data centers will approach 1 trillion kilowatt-hours, of which 40 percent of the additional electricity demand will rely on renewable energy. China has clear scale and technological advantages in solar energy, energy storage, and electrification, and will be better positioned to meet future global demand for green energy and industrial development.

Fourth, China’s industrial development helps improve people’s well-being around the world. The rapid development of Chinese industry has provided the world with a rich supply of high-quality, highly efficient, and cost-effective products, offering consumers in all countries more stable and more diversified choices. Chinese manufacturing has improved living standards, reduced the cost of living, and eased global inflationary pressure. For example, Chinese air conditioners have recently sold well in Europe, helping local people stay cool during heat waves. A report by the European Central Bank estimates that if EU imports from China increase by 10 percent in 2026, the EU’s overall import prices would fall by 1.6 percent. China’s trade and investment have also had a stronger enabling effect on industrialization in developing countries. China has established more than 50,000 enterprises overseas, with outward investment stock exceeding $3 trillion, nearly 90 percent of which is located in developing economies. Through local production, local procurement, local hiring, supporting upstream and downstream industries, and connecting regional supply chains, Chinese-invested enterprises have brought a large number of projects in light industry, textiles, home appliances, and other sectors to fruition, while also driving the development of digital and green sectors. This has strengthened the contribution of exports to local value added in host countries. According to estimates by the Chinese Academy of Sciences, the value added in local exports driven by Chinese-invested enterprises in 24 developing economies increased from $24.4 billion in 2012 to $142.4 billion in 2025, nearly a fivefold increase. The so-called theory that China is squeezing others out is simply a new version of the “China threat” narrative. It is intended to undermine China’s cooperation with countries of the Global South and to shift historical and present responsibilities, rather than to genuinely help developing countries. Facts and data show clearly that China’s industrial development brings the world opportunities, not shocks; empowerment, not threats. It helps developing countries modernize and helps pave the way for their development.

Thank you.

Reporter from Yangguangwang, China National Radio, CMG:

In recent years, unilateralism and protectionism have been on the rise. Some economies have adopted discriminatory measures that undermine fair competition and disrupt global industrial division of labor and cooperation. How do you view the role of market competition in adjusting the global distribution of capacity, and how can the WTO play a better role in maintaining a fair international competitive environment? Thank you.

Yan Dong:

Thank you for your question. I will ask Director-General Han Yong to answer it.

Han Yong:

Thank you for the question. Market competition is an important safeguard for optimizing and adjusting capacity and for promoting healthy industrial development. Historically, every industrial revolution and technological transformation has been accompanied by a rise in capacity in related industries and even, in the short term, situations in which supply exceeded demand. In this process, enterprises invest and expand production in pursuit of profits and market share, while market competition drives them to reduce costs and improve efficiency, bringing technological progress and productivity gains. In that sense, market competition is the most effective mechanism for preventing disorderly expansion of capacity. The role of government should be to maintain competitive order and a fair environment, allowing the market to function more fully so that outdated capacity exits naturally through competition and dynamic balance between supply and demand is ultimately achieved. The WTO’s principle of fair competition and related rules have become widely accepted norms of behavior globally. All parties should uphold fair competition and reduce improper interference with global cooperation in capacity allocation.

China is actively fostering a first-class business environment characterized by fair competition. China has the largest number of market entities in the world. This huge number creates a fully competitive environment in which enterprises face direct challenges and compete on real strength, refining products and services in the survival-of-the-fittest process. McKinsey once described China as “the world’s toughest gym,” a place that has forged highly competitive companies. China’s 15th Five-Year Plan outline makes clear that it will deepen the development of a unified national market, remove barriers in factor acquisition, qualification recognition, tendering and bidding, and government procurement, fully implement national treatment for foreign-invested enterprises, and actively foster a first-class business environment that is market-oriented, law-based, and internationalized.

The multilateral trading system with the WTO at its core and rules at its foundation has played an important role in maintaining the stability of the global economic and trade order. Principles such as free trade, non-discrimination, and fair competition are deeply rooted and together form the underlying logic that keeps the global economic and trade order stable and predictable. At present, the multilateral trading system is under severe impact from unilateralism and protectionism, but multilateralism remains the first choice of the overwhelming majority of WTO members. According to WTO statistics, 72 percent of global trade is still conducted under the most-favored-nation principle. This fully demonstrates the value and resilience of the multilateral trading system’s basic principles, including fair competition. No one wants to return to a jungle world where the strong prey on the weak. In difficult times, international fairness and justice become all the more precious. China will work with all parties to firmly support the multilateral trading system, advance WTO reform, resolutely uphold the WTO’s basic principles and rules, and maintain a transparent, fair, and inclusive global economic and trade environment.

Thank you.

Reporter from the South China Morning Post:

The EU has recently proposed related trade defense instruments, and the United States has launched a Section 301 investigation targeting the capacity issue. How does the Ministry of Commerce comment on this? Given that domestic supply in related areas continues to exceed domestic demand, what measures is the Ministry taking to prevent export spillover effects from triggering more trade restrictions? Thank you.

Yan Dong:

Thank you for your question. Director-General Lin Weilong of the Policy Research Department will answer.

Lin Weilong:

Thank you for your question. As I understand it, your question actually contains three issues: trade measures by Europe and the United States, China’s domestic demand, and the export spillover effect and the resulting global economic imbalance.

Let me first address the question of the EU’s trade defense instruments and the U.S. Section 301 investigation into the capacity issue. China has repeatedly stated its position on both matters. The EU has continued to introduce protectionist measures against China, seriously undermining the confidence of Chinese enterprises in cooperating with Europe. Given the enormous scale of China-EU cooperation, it is inevitable that differences and frictions will arise, but such differences should not become a pretext for fabricating accusations, still less an excuse for setting restrictions, increasing pressure, and affecting practical cooperation. China is willing to properly handle differences and frictions with the EU through dialogue and consultation. As for the U.S. Section 301 investigation on the capacity issue, this is a typical act of unilateralism that seriously undermines the international economic order. The U.S. side cannot narrowly define production capacity exceeding domestic demand as “overcapacity” and attach the label of “excess.” Nor does it have the right, through a Section 301 investigation, to unilaterally determine whether a trading partner has “overcapacity” and to adopt unilateral restrictive measures. China urges the U.S. side to correct its wrong approach and return to the right track of resolving issues through dialogue and consultation. China will closely follow developments, reserve the right to take necessary measures, and firmly safeguard its legitimate rights and interests.

Second, on China’s domestic demand. As our position paper points out, some people claim that “insufficient domestic demand in China has led to overcapacity,” but this does not conform to the facts. China is not only a manufacturing powerhouse, but also a major consumer market. Domestic demand has long been the main engine of China’s economy. From 2013 to 2024, domestic demand contributed an average of 93 percent to China’s economic growth, with consumption and investment contributing 55 percent and 38 percent, respectively, on average. China’s total retail sales of consumer goods rose from 23.8 trillion yuan in 2013 to 50.1 trillion yuan in 2025, more than doubling. Based on World Bank purchasing power parity calculations, China’s retail sales in 2025 were about 1.7 times those of the United States, meaning China is already the world’s largest commodity consumption market in practical terms. For example, in food consumption, China’s Engel coefficient has fallen to 29.8 percent, and per capita protein supply has exceeded 130 grams per day, higher than in many developed countries. In terms of industrial goods consumption, annual per capita purchases of air conditioners, refrigerators, mobile phones, and automobiles have already approached OECD-country levels.

At present, insufficient effective demand in China is a stage-specific phenomenon in the transition from high-speed growth to high-quality development. Over the long term, however, China’s consumption space remains vast and full of potential, and its role as the main engine will become even more prominent. Over the next decade, China’s middle-income group is expected to exceed 800 million people, and per capita GDP will reach the level of moderately developed countries. Structurally, China is accelerating the shift from goods-led consumption toward a pattern in which goods and services are equally important. Service consumption is developing rapidly, and by 2030 it is expected to account for more than half of total consumption. China’s 15th Five-Year Plan contains a dedicated chapter on domestic demand, emphasizing the need to treat demand expansion as a strategic priority, expand effective investment, vigorously boost consumption, implement special initiatives to stimulate consumption, upgrade goods consumption, unlock service-consumption potential, foster new forms of consumption, guide new supply through new demand, create new demand through new supply, promote positive interaction among consumption, investment, supply, and demand, and strive to achieve a higher-level balance between supply and demand.

Third, on global economic imbalance. Global economic imbalances have always existed and are a historical norm. Under the global economic structure and international economic governance system formed after World War II, the world economy has experienced a major imbalance roughly every decade or so, and some of these episodes have even triggered international economic and financial crises. Each new round of imbalance prompts wide international discussion, though the focus has kept changing. Some discussions emphasize market factors such as the savings-investment structure and industrial and supply-chain division of labor; others focus on institutional factors such as the international financial system and macroeconomic policy. The roots of economic imbalance are extremely complex.

In recent years, global economic imbalance has taken on new features. Institutions such as the International Monetary Fund believe that the macroeconomic policies of countries, especially fiscal policy, are a key factor behind current global imbalances. The United States has accumulated a huge debt imbalance and needs to improve its fiscal position; Europe suffers from insufficient investment and needs to raise productivity; China needs to expand domestic demand. That view is only one school of thought, but it does reflect the systemic and complex nature of global economic imbalance. Some people seek to link global economic imbalance with the so-called “overcapacity.” This is simplistic attribution, a deliberate attempt to confuse the issue, and motivated by ulterior motives. By contrast, China’s high-quality capacity has not only met domestic demand but has also made important contributions to global development and played a significant role in promoting global economic balance.

Thank you.

Reporter from Elephant News:

China’s emerging industries such as artificial intelligence, new-energy vehicles, and lithium batteries have developed rapidly, and their export performance has also been strong. What are the sources of China’s industrial strengths and momentum? Thank you.

Yan Dong:

Thank you for your question. I will ask Mr. He Shaojun from the Department of Foreign Trade to answer.

He Shaojun:

Thank you. We have also noted that in recent years China’s electric vehicles, lithium batteries, and artificial intelligence industries have developed rapidly and posted strong export growth. According to customs statistics, in the first half of this year, China’s exports of electric vehicles and lithium batteries increased by 68.7 percent and 37.6 percent, respectively. AI-related products such as industrial robots and 3D printers also performed strongly, rising by 18.6 percent and 109.3 percent, and have become new calling cards for China’s foreign trade. Overall, this is the result of multiple factors, including a solid industrial foundation, market-driven iteration, enterprise innovation, and international openness and cooperation. These can be summarized as four major strengths.

First is the strength of a complete industrial system. China has now established a comprehensive and efficient new-energy vehicle industrial chain covering basic materials, components, complete vehicles, and manufacturing equipment. Its supply-chain responsiveness, cost control, and delivery speed are globally leading. In the Yangtze River Delta, through coordinated industrial-cluster development, a new-energy vehicle manufacturer can obtain needed supporting components within a four-hour drive.

Second is the strength of a super-sized market. China has the world’s largest consumer market, which provides a proving ground and training ground for new technologies, products, and services. New-energy vehicle sales have ranked first in the world for 11 consecutive years, and fierce domestic competition has greatly improved product technology. Massive user demand and diverse application scenarios have accelerated AI technology iteration, enabling rapid verification of technical feasibility and spreading R&D costs across a larger base.

Third is the strength of innovation-driven development. China has firmly pursued innovation-driven development, using scientific and technological innovation to lead industrial development and continuously enhance industrial competitiveness. National R&D intensity rose from 1.91 percent in 2012 to 2.8 percent in 2025. International patent applications have ranked first in the world for seven consecutive years, and China accounts for 60 percent of global AI patents. Technological breakthroughs have been achieved in areas such as power batteries and general-purpose large models. By 2025, the energy density of power batteries had increased by more than 50 percent compared with 2018, while production costs had fallen by more than 60 percent. Domestic large models such as DeepSeek and Qwen have risen collectively.

Fourth is the strength of openness and cooperation. At present, the global energy system is undergoing deep transformation, and artificial intelligence is developing rapidly. Economic and trade cooperation in related fields has enormous potential. According to institutional estimates, by 2030 the global market size for products and technologies such as electric vehicles, solar energy, and wind power will reach $2.1 trillion, while the AI data-center market is expected to grow by 45 percent annually over the next five years. The market outlook is broad. Guided by the principle of mutual benefit and win-win cooperation, China advances international cooperation and supports capable electric-vehicle and lithium-battery enterprises in making rational and orderly cross-border arrangements for industrial and supply chains, thereby empowering green transition and industrial upgrading in other countries.

The development of China’s modern industries and the enhancement of its foreign-trade competitiveness are driven by innovation and by the continuous deepening of reform. Going forward, we will continue to advance high-quality trade development, further support enterprises in related industries in integrating more deeply into global industrial and supply-chain systems, and inject more vitality into global digital, intelligent, and green transformation.

Thank you.

Jia Huili:

Thank you. We now have time for one last question.

Reporter from CCTV, China Media Group:

Under the current volatile international situation, maintaining stable and smooth global industrial and supply chains faces many challenges and requires the joint efforts of all countries moving in the same direction. What suggestions does the Ministry of Commerce have for better promoting an open and inclusive global industrial and supply chain cooperation landscape? Thank you.

Yan Dong:

Thank you for your question. I will answer this one.

At present, with the international situation turbulent and protectionism on the rise, safeguarding the stability and smooth functioning of global industrial and supply chains is in the interests of all parties and is also an urgent expectation of the international community. President Xi Jinping has pointed out that economies advance together when connected and all fall behind when closed off. China is willing to work with all parties to uphold the global free-trade system, preserve an international environment of openness and cooperation, and jointly promote the building of an open and inclusive global industrial and supply chain cooperation landscape.

We must uphold mutual benefit and win-win outcomes and make the global development pie bigger. When the pie grows larger, conflicts over distribution become smaller. All parties should jointly seize the opportunities created by a new round of industrial revolution and technological transformation, strengthen international cooperation in areas such as green and low-carbon development, artificial intelligence, and bio-manufacturing, and use shared technological dividends to break through growth bottlenecks and continuously inject new momentum into the global economy. All parties should also pay greater attention to the “real imbalances” between developed and developing economies, strengthen industrial cooperation as well as trade and investment cooperation, help more developing countries and regions integrate into the international division of labor, accelerate industrialization and modernization, and open up a new blue ocean for global industrial cooperation.

We must uphold openness and connectivity and promote sound circulation in global industrial and supply chains. The division of labor and cooperation in global industrial and supply chains did not take shape overnight, and forced intervention by human hands will only backfire. All parties should continue expanding market openness, promote trade and investment liberalization and facilitation, improve the efficiency of allocating factors and resources, and jointly create and maintain an open global innovation ecosystem so that international industrial cooperation can deliver greater benefits. All parties should oppose the politicization and over-securitization of economic issues, respect market rules, reduce barriers to trade, lower obstacles to investment cooperation, promote full market competition, stimulate business vitality, and make the flow of factors and resources and the distribution of industrial division of labor more rational and orderly.

We must uphold policy coordination and foster a stable and predictable environment for cooperation. The world economy is highly interconnected, and national industrial policies have obvious spillover effects, so communication and coordination should be strengthened, especially among major powers, which should set an example. All parties should adhere to consultation on an equal footing and properly manage differences, abandon unilateralism and protectionism, and oppose discriminatory and exclusionary practices. All parties should strengthen bilateral and multilateral dialogue on industrial policies, uphold openness and transparency, and deepen exchanges and discussions on industrial policy under the WTO framework, taking more coordinated and effective measures to better pool the combined forces of global economic growth.

We must uphold multilateralism and build a fairer and more equitable international economic order. The multilateral trading system with the WTO at its core is the cornerstone of economic globalization and international trade. In economic and trade exchanges, all parties should uphold equality and mutual benefit, respect each other’s stage of development and national conditions, and strive through fair competition to run faster themselves rather than trip others up, while jointly resisting bullying by the strong against the weak. All parties should adhere to genuine multilateralism, uphold the WTO’s basic principles and rules, advance WTO reform in step with the times, safeguard the WTO’s authority and effectiveness, and make better use of multilateral and regional cooperation mechanisms such as the G20, BRICS, and APEC, so as to jointly defend fairness and justice and make the global economic governance system more just and equitable.

Thank you.

Closing

Jia Huili:

That concludes today’s press conference. Thank you to all the speakers and to all the journalists for participating. Goodbye.

Kimi K3 Open Day

Thank you for your patience.

Today is Kimi K3 Open Day. We are releasing the Kimi K3 model weights, the technical report, and the key infrastructure technologies that support Kimi K3 model training: MoonEP, FlashKDA, and AgentEnv. We hope this will accelerate the deployment and adoption of frontier intelligence and advance AGI research.

Kimi K3 Weights Released

Kimi K3 is our most capable model. It is a 2.8-trillion-parameter Mixture-of-Experts model with native visual understanding and support for a 1-million-token context window.

Kimi K3 has roughly three times as many parameters as Kimi K2.5. Scaling, however, is not merely the accumulation of parameters. Under constrained compute conditions, we achieved a 2.5× improvement in scaling efficiency through a set of technical innovations, including Kimi Delta Attention, Attention Residuals, and MoonEP. In compute-optimal terms, each unit of compute now produces roughly 2.5 times as much intelligence as before.

Everyone can now download and deploy the Kimi K3 model, whether for internal research and development or for integration into end-user products. For other use cases, please see the Kimi K3 License.

Download Kimi K3 model weights

Kimi K3 Technical Report Published

Alongside the Kimi K3 model weights, we are also making public our model training methods. The Kimi K3 technical report is now available.

The technical report covers the following details:

• KDA + AttnRes: KDA and Gated MLA are mixed at a 3:1 ratio to enable efficient long-context modeling, while block-level attention residuals improve cross-layer information flow.

• Stable LatentMoE: Each token activates 16 out of 896 routed experts. SiTU-GLU and Quantile Balancing maintain training stability under extremely high sparsity.

• MoonViT-V2: The visual encoder is trained from scratch using next-token prediction, without contrastive pretraining. It reaches the baseline performance of SigLIP initialization while providing a more stable optimization process.

• Post-training and evaluation: The report covers large-scale task synthesis across general reasoning, general agents, and coding agents; reinforcement learning infrastructure for million-token contexts; and full evaluation results from nearly 20 internal benchmarks.

These points are only an overview. Detailed discussions and ablation experiments are included in the technical report.

Download the Kimi K3 technical report

Key Kimi K3 Infrastructure Technologies Open-Sourced

Model capability depends on stable training systems at the infrastructure layer. Today, we introduce three infrastructure technologies that support Kimi K3 training: MoonEP, FlashKDA, and AgentEnv. They cover key links from high-performance communication and high-performance kernels to distributed RL environments. FlashKDA had already been open-sourced; MoonEP and AgentEnv are being officially open-sourced with this release.

• MoonEP: MoonEP is a high-performance communication library designed for extremely large, fine-grained MoE models. It enables expert-parallel communication to maintain high efficiency even under imbalance.

• FlashKDA: FlashKDA is our high-performance kernel implementation of Kimi Delta Attention. On NVIDIA H20, compared with the flash-linear-attention baseline, its prefill speed is 1.72–2.22× faster, and it can be used directly as a replacement backend for flash-linear-attention.

• AgentEnv: AgentEnv is a sandbox system developed jointly with KVCache.ai for running agent environments at scale. It provides high-fidelity, strongly isolated sandboxes for Kimi K3 post-training, with flexible support for fast snapshotting, restoration, and forking. It is designed for massively parallel agent workflows and training tasks.

These three technologies are key to Kimi K3’s training efficiency and stability. We are opening them up in the hope that they can also help you train next-generation models.

MoonEP GitHub repository

FlashKDA GitHub repository

AgentEnv GitHub repository

Why We Choose Openness

We firmly believe in the value of open-weight models. They lower the barrier to accessing intelligence, promote innovation, and give users greater control over their data, privacy protection, and ownership. Over the past few weeks, we have been pleased to hear support from many members of the AI community and from technology leaders who share our vision.

We believe that for a technology with far-reaching impact like AGI, a broad and open ecosystem is the most suitable foundation. We will continue to contribute to that ecosystem.

Quick Start

• Download Kimi K3 model weights

• Read the Kimi K3 technical report

• Learn about MoonEP

• Learn about FlashKDA

• Learn about AgentEnv

• Learn about the Kimi K3 model

GitHub Repository Content

Repository Overview

MoonshotAI/Kimi-K3 is titled “Open Frontier Intelligence.” The repository is public and includes an assets directory, LICENSE, README.md, and k3_tech_report.pdf. The repository page links to the Kimi chat site, Moonshot AI homepage, Hugging Face organization, Twitter/X account, Discord, ModelScope organization, the Kimi K3 license, the technical blog, and the full report.

1. Model Introduction

Kimi K3 is an open-weight, native multimodal agentic model and Moonshot AI’s most capable model to date. It is a 2.8T-parameter model built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), with native vision capabilities and a 1-million-token context window. It is described by the repository as the world’s first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.

Key Features

• New Architecture: Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), and scales up MoE sparsity with a Stable LatentMoE framework that activates 16 out of 896 experts, yielding an approximate 2.5× improvement in overall scaling efficiency over Kimi K2.

• Long-Horizon Coding: Operating with minimal human oversight, Kimi K3 sustains long engineering sessions, navigates massive repositories, and orchestrates terminal tools, from GPU kernel optimization and compiler development to vision-in-the-loop game development, CAD, and chip design.

• Agentic Knowledge Work: Kimi K3 advances end-to-end knowledge work, producing deep research with interactive visualizations, widgets and dashboards, motion design, and video editing, powered by its native multimodal architecture.

• Native Multimodality & Long Context: Kimi K3 understands text, images, and video within the same model, and supports a 1-million-token context window.

• Open Frontier Weights: Moonshot AI releases the full Kimi K3 model weights under the Kimi K3 License, making frontier intelligence openly available for research, deployment, and further innovation.

2. Model Summary

3. Evaluation Results

3. Benchmark | Kimi K3 | Claude Fable 5 | GPT-5.6 Sol | Claude Opus 4.8 | GPT-5.5 | GLM-5.2
     <tr><td>GPQA Diamond</td><td>93.5</td><td>92.6</td><td>94.1</td><td>91.0</td><td>93.5</td><td>91.2</td></tr>  
     <tr><td>CritPt</td><td>23.4</td><td>28.6</td><td>32.3</td><td>20.9</td><td>27.1</td><td>20.9</td></tr>  
     <tr><td>AA-LCR</td><td>74.7</td><td>70.0</td><td>73.7</td><td>67.7</td><td>74.3</td><td>71.3</td></tr>  
     <tr><td>HLE-Full</td><td>43.5 / 56.0</td><td>53.3 / 63.0</td><td>44.5 / 58.0</td><td>49.8 / 57.9</td><td>41.4 / 52.2</td><td>—</td></tr>  
     <tr><td>DeepSWE</td><td>67.5</td><td>70.0</td><td>73.0</td><td>59.0</td><td>67.0</td><td>46.2</td></tr>  
     <tr><td>ProgramBench</td><td>77.8</td><td>76.8</td><td>77.6</td><td>71.9</td><td>70.8</td><td>63.7</td></tr>  
     <tr><td>Terminal-Bench 2.1</td><td>88.3</td><td>88.0</td><td>88.8</td><td>84.6</td><td>83.4</td><td>82.7</td></tr>  
     <tr><td>FrontierSWE</td><td>81.2</td><td>86.6</td><td>71.3</td><td>66.7</td><td>64.9</td><td>67.3</td></tr>  
     <tr><td>SWE-Marathon</td><td>42.0</td><td>35.0</td><td>39.0</td><td>40.0</td><td>14.0</td><td>13.0</td></tr>  
     <tr><td>PostTrainBench</td><td>36.6</td><td>41.4</td><td>34.6</td><td>34.1</td><td>28.4</td><td>34.3</td></tr>  
     <tr><td>MLS-Bench-Lite</td><td>48.3</td><td>49.9</td><td>46.2</td><td>42.8</td><td>35.5</td><td>40.4</td></tr>  
     <tr><td>SciCode</td><td>58.7</td><td>60.2</td><td>56.1</td><td>53.5</td><td>56.1</td><td>50.5</td></tr>  
     <tr><td>Kimi Code Bench 2.0</td><td>72.9</td><td>76.9</td><td>64.8</td><td>71.7</td><td>69.0</td><td>64.2</td></tr>  
     <tr><td>BrowseComp</td><td>91.2</td><td>88.0</td><td>90.4</td><td>84.3</td><td>84.4</td><td>—</td></tr>  
     <tr><td>DeepSearchQA F1</td><td>95.0</td><td>94.2</td><td>—</td><td>93.1</td><td>—</td><td>—</td></tr>  
     <tr><td>ResearchRubrics</td><td>76.2</td><td>—</td><td>73.8</td><td>73.5</td><td>64.0</td><td>71.1</td></tr>  
     <tr><td>GDPval-AA v2 Elo</td><td>1686</td><td>1747</td><td>1736</td><td>1593</td><td>1491</td><td>1510</td></tr>  
     <tr><td>Toolathlon-Verified</td><td>76.5</td><td>77.9</td><td>74.9</td><td>76.2</td><td>73.5</td><td>59.9</td></tr>  
     <tr><td>MCPMark-Verified</td><td>94.5</td><td>87.4</td><td>92.9</td><td>76.4</td><td>92.9</td><td>—</td></tr>  
     <tr><td>MCP-Atlas</td><td>84.2</td><td>84.7</td><td>83.6</td><td>83.6</td><td>82.8</td><td>82.6</td></tr>  
     <tr><td>AutomationBench</td><td>30.8</td><td>29.1</td><td>29.7</td><td>27.2</td><td>22.7</td><td>12.9</td></tr>  
     <tr><td>JobBench</td><td>54.3</td><td>57.4</td><td>45.4</td><td>48.4</td><td>38.3</td><td>43.4</td></tr>  
     <tr><td>AA-Briefcase Elo</td><td>1548</td><td>1583</td><td>1495</td><td>1354</td><td>1158</td><td>1260</td></tr>  
     <tr><td>Agents' Last Exam</td><td>28.3</td><td>25.7†</td><td>29.6</td><td>27.0</td><td>26.6</td><td>20.4</td></tr>  
     <tr><td>APEX-Agents</td><td>41.0</td><td>43.3</td><td>39.9</td><td>39.4</td><td>38.5</td><td>35.6</td></tr>  
     <tr><td>OfficeQA Pro</td><td>63.3</td><td>69.9</td><td>63.2</td><td>63.9</td><td>60.9</td><td>41.4</td></tr>  
     <tr><td>SpreadsheetBench 2</td><td>34.8</td><td>34.7</td><td>32.4</td><td>31.6</td><td>29.1</td><td>28.1</td></tr>  
     <tr><td>OSWorld-Verified</td><td>84.8</td><td>85.0</td><td>83.0</td><td>83.4</td><td>79.0</td><td>—</td></tr>  
     <tr><td>OSWorld 2.0</td><td>58.3</td><td>66.1</td><td>62.6</td><td>55.7</td><td>49.5</td><td>—</td></tr>  
     <tr><td>SaaS-Bench</td><td>60.1</td><td>—</td><td>61.4</td><td>56.1</td><td>43.8</td><td>—</td></tr>  
     <tr><td>τ³-Banking</td><td>33.4</td><td>26.8</td><td>33.0</td><td>27.6</td><td>31.3</td><td>26.8</td></tr>  
     <tr><td>Harvey Lab-AA</td><td>94.6</td><td>93.6</td><td>87.2</td><td>91.1</td><td>86.3</td><td>91.0</td></tr>  
     <tr><td>CorpFin v2</td><td>71.6</td><td>71.8</td><td>64.4</td><td>66.7</td><td>68.4</td><td>66.1</td></tr>  
     <tr><td>Finance Agent v2</td><td>54.4</td><td>56.3</td><td>53.8</td><td>53.9</td><td>51.8</td><td>49.7</td></tr>  
     <tr><td>Legal Research Bench</td><td>44.2</td><td>49.5</td><td>48.1</td><td>43.8</td><td>40.4</td><td>31.3</td></tr>  
     <tr><td>WorldVQA ForceAnswer</td><td>51.0</td><td>56.7</td><td>41.8</td><td>39.1</td><td>38.5</td><td>—</td></tr>  
     <tr><td>OmniDocBench</td><td>91.1</td><td>89.8</td><td>85.8</td><td>87.9</td><td>89.4</td><td>—</td></tr>  
     <tr><td>PerceptionBench</td><td>58.5</td><td>57.2</td><td>59.7</td><td>47.2</td><td>55.8</td><td>—</td></tr>  
     <tr><td>Video-MME w. sub</td><td>90.0</td><td>—</td><td>89.5</td><td>86.0</td><td>89.3</td><td>—</td></tr>  
     <tr><td>MMVU</td><td>82.1</td><td>—</td><td>81.2</td><td>79.2</td><td>81.7</td><td>—</td></tr>  
     <tr><td>BabyVision w/ python</td><td>85.7</td><td>90.5</td><td>88.9</td><td>81.2</td><td>83.6</td><td>—</td></tr>  
     <tr><td>MMMU-Pro</td><td>81.6 / 83.4</td><td>81.2 / 86.5</td><td>83.0 / 84.6</td><td>78.9 / 82.7</td><td>81.2 / 83.2</td><td>—</td></tr>  
     <tr><td>CharXiv RQ</td><td>84.8 / 91.3</td><td>88.9 / 93.5</td><td>84.6 / 89.1</td><td>80.5 / 89.9</td><td>84.1 / 89.0</td><td>—</td></tr>  
     <tr><td>MathVision</td><td>94.3 / 97.8</td><td>94.8 / 98.6</td><td>95.8 / 97.8</td><td>86.7 / 97.1</td><td>92.2 / 96.8</td><td>—</td></tr>  
     <tr><td>ZeroBench pass@5</td><td>23.0 / 41.0</td><td>23.0 / 46.0</td><td>17.0 / 35.0</td><td>17.0 / 34.0</td><td>22.0 / 41.0</td><td>—</td></tr>  
   

The repository footnotes state that all Kimi K3 results are obtained with reasoning effort set to max and temperature set to 1.0. For single-step tasks such as GPQA Diamond, HLE-Full, and vision benchmarks without tools, top-p is set to 0.95. For agentic tasks, top-p is set to 1.0. For HLE-Full, MMMU-Pro, CharXiv (RQ), MathVision, and ZeroBench, each cell reports scores without and with tool augmentation in that order.

Evaluation Footnotes

Reasoning and knowledge benchmark scores for CritPt and AA-LCR are cited from Artificial Analysis as of July 23, 2026.

For coding benchmarks, DeepSWE is evaluated with the Kimi Code harness. GLM-5.2 is taken from the GLM-5.2 release blog, while remaining DeepSWE scores are from the official DeepSWE leaderboard; Kimi K3 attains 67.3 with the mini-SWE-agent harness. Terminal-Bench 2.1 reports the best score across harnesses for other models. ProgramBench uses the Kimi Code harness for Kimi K3 and Vals AI for other non-GLM scores. SWE-Marathon is based on an H20-calibrated branch of official tasks as of July 9, 2026. FrontierSWE dominance scores are recomputed from raw scores using the official evaluation script and are current as of July 16, 2026. PostTrainBench scores for selected models are adopted from the official results, while Kimi K3, Claude Fable 5, and GPT-5.6 Sol are evaluated with Harbor at maximum reasoning effort over three runs on H20 GPUs.

For agentic benchmarks, OfficeQA Pro provides each test case with the entire PDF corpus as images and no machine-readable text. MCP-Atlas uses the 500-task public subset with a 100-turn limit and Gemini 3.1 Pro as the judge. AutomationBench uses the 600-task public subset. BrowseComp adopts a context-compaction strategy triggered at 300K tokens; with the full 1M-token context window and no context management, Kimi K3 scores 90.4. Several benchmark scores are cited from Artificial Analysis, APEX-Agents, Vals AI, and Agents’ Last Exam leaderboards as of July 23, 2026.

For multimodal benchmarks, all scores except ZeroBench are averaged over three runs. ZeroBench follows the official setting and is run five times. MMMU-Pro preserves the original input order and prepends images to the text input. PerceptionBench is described as an in-house benchmark focused on atomic visual perception capabilities.

4. Native MXFP4 Quantization

Kimi K3 applies quantization-aware training from the SFT stage onward, using MXFP4 weights with MXFP8 activations for broad hardware compatibility.

5. Deployment

Kimi K3’s API can be accessed at platform.kimi.ai by selecting kimi-k3. Moonshot AI provides OpenAI-compatible and Anthropic-compatible APIs. The repository recommends the following inference engines:

• vLLM, with recipes

• SGLang, with cookbook

• TokenSpeed, with recipes

6. Model Usage

Kimi K3 always has thinking enabled and returns reasoning_content. Thinking effort is configured through the top-level reasoning_effort request field, which supports low, high, and max; the default is max.

Kimi K3 was trained in preserved thinking history mode. For multi-turn conversations and tool calls, Kimi K3 requires the complete assistant message returned by the API to be passed back into messages unchanged, including reasoning_content and tool_calls, not only content.

import openai

def chat_with_preserved_thinking(client: openai.OpenAI, model_name: str):
    messages = [
        {
            "role": "user",
            "content": "Tell me three random numbers."
        },
        {
            "role": "assistant",
            "reasoning_content": "I'll start by listing five numbers: 473, 921, 235, 215, 222, and I'll tell you the first three.",
            "content": "473, 921, 235"
        },
        {
            "role": "user",
            "content": "What are the other two numbers you have in mind?"
        }
    ]

    response = client.chat.completions.create(
        model=model_name,
        messages=messages,
        stream=False,
        max_tokens=4096,
        reasoning_effort="max",
    )
    # the assistant should mention 215 and 222 that appear in the prior reasoning content
    print(f"response: {response.choices[0].message.reasoning}")
    return response.choices[0].message.content

For full guides and examples covering vision input, structured output, partial mode, tool choice, dynamic tool loading, and context caching, see the Kimi K3 Quickstart and Thinking Effort.

Coding Agent Framework

Kimi K3 works best with Kimi Code CLI as its agent framework. The repository invites users to run Kimi Code in the terminal and select Kimi K3 using the /model command.

7. License

Both the code repository and the model weights are released under the Kimi K3 License.

8. Contact

For questions, the repository lists support@moonshot.ai.

George Chen, Partner at The Asia Group, has spent the past few days attending the APEC Digital Ministers' High-Level Meeting in Chengdu. According to him, the U.S. delegation has already raised the issue of model distillation with the Chinese delegation during closed-door discussions.

In closed‑door meetings, the U.S. delegation pressed Chinese officials on “industrial‑level distillation”, citing the new release of Kimi K3 as an example. U.S. officials also claimed training new models on the outputs of existing ones could undermine intellectual property protections. CAC officials strongly pushed back, insisting distillation was legitimate and distinct from theft.

After largely remaining silent throughout the debate for several days, China offered its first comprehensive response today.

At a regular press briefing, a spokesperson for China’s Ministry of Commerce (MOFCOM) delivered a lengthy and carefully worded statement.

MOFCOM first framed the U.S. allegations as an attempt to politicize and weaponize technology and trade issues, “stigmatize Chinese companies,” and “label” China without evidence. It further described Washington’s approach as a form of “AI hegemonism.”

The ministry also questioned the factual basis of the U.S. accusations. It noted that the release of Kimi K3 (without naming the company directly, referring instead to “the relevant Chinese enterprise”) came very close to the release of comparable U.S. frontier models. It further argued that Chinese AI models have already reached—or in some areas surpassed—the global frontier, citing front-end coding capabilities as one example of China’s growing technological competitiveness.

MOFCOM then turned the argument around, stating that many American AI companies have themselves distilled Chinese models during research and training. It also pointed to recent statements by “some large U.S. multinational companies”—an apparent reference to firms such as NVIDIA—which have argued that model distillation is a widely used industry practice and that American companies should continue to have access to Chinese open-source AI models. In particular, MOFCOM cited the recent open letter signed by dozens of U.S. companies supporting open-weight models as evidence that significant parts of the American AI industry disagree with the current U.S. government’s approach.

Finally, MOFCOM urged Washington to “stop smearing Chinese companies and threatening sanctions.” It also warned that if U.S. actions were to “substantially harm” China’s legitimate interests, Beijing would take “all necessary measures” to safeguard its lawful rights and interests.

The statement was widely republished by Xinhua and other state media outlets, as well as by major Chinese online commentators, quickly becoming one of today’s biggest AI policy stories in China.

Almost simultaneously, Yuyuantantian, a semi-official media account affiliated with CCTV, published a feature commentary titled “Open Models Can Break Western Bias.” The commentary opened with the recent incident in which Hugging Face was autonomously compromised by OpenAI’s models and ultimately relied on Zhipu AI’s GLM-5.2 to resolve the problem. It argued that while the West has long attacked the security of open-weight models, the Hugging Face incident has forced people to reconsider who should define AI safety, risk, and the future of artificial intelligence.

The commentary argued that the United States and the West have been fixated on using “security” as the primary argument against open-weight models, mainly to create justification for future restrictions and sanctions.

Once a technology is labeled a “security threat,” virtually any form of government intervention can be presented as legitimate. Export controls, licensing regimes, and even excluding specific countries from global supply chains can all be framed as measures taken “for the benefit of humanity.”

The article also reviewed Western media coverage over the past several years regarding the safety of open and proprietary models, arguing that these narratives have followed a clear pattern. According to the commentary, they have consistently promoted the following storyline:

Open models mean uncontrolled capabilities and the uncontrolled spread of risk; proprietary models represent responsible governance and order; and when Chinese models enter the global open-source ecosystem, their progress is simply attributed to leveraging Western technology.

Yuyuantantian argued that this narrative is overly absolutist.

There is no such thing as a perfect technological pathway. The existence of problems should not automatically lead to the conclusion that a particular technological approach is destined to fail. More importantly, what matters is who can better identify problems and adapt under pressure.

According to Yuyuantantian, the open-weight ecosystem has already demonstrated significant advantages in at least three respects, and the world does not have to choose between open and proprietary models as if they were mutually exclusive. The recent collective opposition by American companies to banning open-weight models illustrates precisely this point.

First, identifying problems and solving them. Research institutions, enterprise users, and independent developers can inspect model behavior, identify vulnerabilities, develop safeguards, and continuously improve models without waiting for a single vendor to issue updates.

Second, real-world deployment accelerates model improvement. Once open-weight models are deployed in practice, their performance across different scenarios is continuously tested. The more they are used, the clearer the direction for future technical improvements becomes.

Third, specialization is beginning to emerge within the ecosystem. Companies are increasingly deploying open and proprietary models for different categories of tasks. NVIDIA and others have begun offering dedicated large language model routing solutions that dynamically select different models depending on the task.

The commentary argues that the open-weight approach represents a viable path for the continuous evolution of AI and for benefiting users around the world. The future of artificial intelligence depends on whether it can accommodate more choices. Now that both open and proprietary models have become part of the AI industry, the question is no longer choosing one over the other, but rather establishing a set of rules that allows both approaches to maximize their respective strengths while respecting their respective boundaries.

The commentary presents this as an entry point for understanding China’s approach to AI governance. From the Global Development Initiative, the Global Security Initiative, the Global Civilization Initiative, to the Global AI Governance Initiative, China has consistently advocated respecting each country’s own choices, addressing shared challenges through cooperation, and promoting common development through the provision of global public goods. According to the article, this philosophy has now been extended to artificial intelligence.

At a forum organized by the Chinese People’s Association for Public Diplomacy, one expert put it quite directly: China has already provided many public goods to the international community. In the field of AI, countries should simply “use whichever models work best,” and China is willing to strengthen coordination with governments around the world.

According to the commentary, Yuyuantantian also spoke with “industry insiders” and put forward several proposals. Taken together, these proposals aim to shift AI governance away from the binary debate over open versus proprietary models toward a framework of capability- and risk-based governance: opening what should be open, restricting what should be restricted, determining infringement based on evidence rather than assumptions based on a model’s country of origin, and establishing mechanisms for model safety evaluation, vulnerability disclosure, major incident reporting, and international cooperation while preserving innovation and controlling high-risk capabilities.

First, define clear boundaries for openness.

The key question going forward is no longer whether to support open or proprietary models, but rather: Which capabilities should be openly released? Which capabilities should be released only under controlled conditions? Which capabilities require access controls? And after a model is released, who should be responsible for vulnerability disclosure, major incident reporting, and governance of downstream derivatives?

According to an individual who has long studied AI policy and was quoted by Yuyuantantian, basic capabilities can be openly released, some frontier capabilities can be released under controlled conditions, commercial services can remain proprietary, while high-risk capabilities should be subject to access controls and safety evaluations. Before release, dangerous capabilities should be assessed. After release, mechanisms should exist for vulnerability disclosure, major incident reporting, and governance of derivative models.

China supports openness, but this does not mean advocating the unconditional dissemination of every capability. AI governance should begin by assessing what capabilities a model possesses and what risks it may create, rather than focusing on which country developed the model.

Second, define clear boundaries for so-called “infringement.”

Whether infringement has occurred should be determined on the basis of verifiable evidence. One should not simply infer that a model must have “stolen” American technology merely because its capabilities have improved or because it was developed in China.

Third, establish security cooperation mechanisms for the open ecosystem.

AI risks can spread across platforms and across national borders. Countries should promote compatible evaluation standards and establish mechanisms for reporting major security incidents and coordinating vulnerability responses.

Ultimately, the goal is to leave room for innovation, establish clear boundaries for risk, and preserve choices for all countries.

The real dividing line in the next phase of AI governance will not be between open and proprietary models, but between indiscriminate restrictions and risk-based governance.

Whoever succeeds in making AI more broadly accessible while keeping high-risk capabilities under effective control is more likely to define the next stage of artificial intelligence.

The debate unfolding in the United States over open-weight models has also prompted some Chinese observers to reflect on a broader question: how should China manage the relationship between open and proprietary models going forward?

Some observers argue that the current U.S. debate offers valuable lessons for China as well. While China has consistently promoted openness, accessibility, and shared AI development internationally, open-weight models are not the entirety of China’s AI ecosystem. China also has a number of companies making serious long-term investments in proprietary frontier models.

According to them, for large Chinese technology companies, the commercial logic behind open-weight models has become increasingly clear. The strategy is to offer models for free or at very low cost, expand adoption among developers and enterprises, and monetize downstream through cloud computing, inference services, enterprise deployment, applications, developer tools, and hardware. Many industry observers, for example, view Qwen primarily as a customer acquisition channel for Alibaba Cloud. According to Alibaba, by early 2026 the Qwen family had surpassed one billion cumulative downloads on Hugging Face. During the same period, Alibaba Cloud’s external revenue grew by 40% year over year, while AI-related products continued to post triple-digit growth.

They also argued that the picture is very different for AI startups that have models but lack their own cloud platform, application layer, or customer distribution channels. For these companies, a purely open-weight business model may prove difficult to sustain over the long term. Training a frontier model can require investments of billions of dollars. Yet once the model weights are released, competitors can download them, build their own APIs, quantize and fine-tune the models, integrate them into their own cloud platforms, and serve customers at lower prices—all without paying anything to the original model developer. The result is that the model company bears the overwhelming R&D costs, while much of the commercial value accrues to cloud providers, application developers, and distribution channels.

This is not a major problem for Alibaba. Alibaba owns not only the model, but also the cloud platform, enterprise customer relationships, and an integrated commercial ecosystem. The model itself drives demand for higher-value cloud and enterprise services. For startups whose only asset is the model, however, this could become a fundamental commercial dilemma.

As more Chinese models are released as open weights or offered at extremely low prices, enterprises may become increasingly unwilling to pay several times more for a proprietary API that is only marginally better. Model capabilities are likely to become commoditized much more rapidly. API prices may continue to fall, the pricing power of standalone model developers may weaken, and value creation may gradually shift toward cloud platforms, data, applications, enterprise services, and customer relationships.

This suggests that the business models of Chinese AI companies may also evolve over time. Open-weight development is likely to continue because it is no longer merely a technical philosophy. It has become an important competitive strategy for attracting global developers, challenging the platform advantages of U.S. firms, and expanding the adoption of China’s AI technology stack. At the same time, however, the most expensive, most advanced, most commercially valuable, and most sensitive capabilities may no longer be fully open. Instead, they may increasingly remain proprietary and be monetized through APIs, cloud services, enterprise solutions, and vertical applications.

If this trajectory continues, China and the United States may ultimately converge toward a remarkably similar industrial structure: open foundation models competing for developers, ecosystem adoption, and market share, while proprietary frontier models maintain a performance lead and generate revenue through cloud services, computing infrastructure, enterprise offerings, and applications. This is, in fact, remarkably close to the long-term vision Jensen Huang has described.

If the industry evolves in this direction, the Chinese government will face not only industrial policy questions but also regulatory ones.

Some observers therefore argue that China needs to think carefully, sooner rather than later, about how to regulate the relationship between open and proprietary models, including policy guidance, industrial support, and technology export controls.

For example, global media have recently reported that China is considering export controls on advanced domestic AI models in order to prevent cutting-edge AI technologies from flowing overseas. If those reports prove accurate, policymakers would at least face two possible approaches.

The first would be to subject both open and proprietary models to the same export control regime. This would provide regulatory consistency. However, because open-weight models are inherently designed to be copied and distributed, whether such controls could be effectively enforced in practice is another question. More importantly, countries across the Global South might begin to question whether China’s long-standing commitment to openness, accessibility, and shared technological development is changing, making them less willing to adopt China’s AI technology stack.

The second approach would be to focus export controls primarily on proprietary models while allowing open-weight models to remain relatively unrestricted. Under this approach, a model such as Kimi K3, because it is released with open weights, could continue to publish its weights on major international open-source platforms and be freely deployed overseas without requiring export approval. By contrast, a proprietary model—even one that is somewhat less capable than K3—might be required to obtain government approval before its weights could be transferred abroad or deployed internationally.

Such a regulatory framework would give open-weight models a natural competitive advantage in perhaps the most important aspect of global expansion: cross-border weight distribution and overseas deployment. That advantage would not arise from superior technology, but rather from differences in government regulatory treatment.

Beijing responded the following day. On 24 July, China’s Ministry of Commerce (MOFCOM) issued Announcement No. 30 of 2026, adding 14 EU entities to China’s Export Control List with immediate effect. The announcement itself is terse. It prohibits Chinese exporters from supplying controlled dual-use items to the named entities; bars overseas parties from transferring Chinese-origin dual-use items to them; orders any transactions already under way to stop immediately; and leaves open a licensing channel for exceptional cases.

In its public explanation, MOFCOM described the measure as a response to what it called the EU’s “egregious behaviour”. The ministry cited the Export Control Law and the Regulations on Export Control of Dual-Use Items as its legal basis.

The speed, scale and composition of the Chinese list suggest that several considerations were at work. They do not all operate on the same timetable, and some are more clearly supported by the available evidence than others.

The 14 entities are spread across eight EU member states:

● Germany: Rheinmetall, Sindlhauser Materials and Antraco Chemie

● France: InPACT, III-V Lab and Cavok UAS

● Italy: Lafert and Garnet

● Poland: Vigo Photonics and Wroclaw University of Science and Technology

● The Netherlands: IHC Merwede

● The Czech Republic: TATRA TRUCKS

● Bulgaria: Opticoelectron Group

● Lithuania: Ekspla

All eight countries have made national trade, financial or defence-industrial commitments in support of Ukraine. That common feature is relevant, although it does not by itself establish why each entity was selected.

By sector, the list is weighted towards defence and dual-use manufacturing, photonics, infrared systems, lasers and specialist materials. Four entities can broadly be described as defence primes or systems integrators: Rheinmetall, TATRA, Cavok UAS and Opticoelectron. Four operate in photonics, infrared detection or lasers: III-V Lab, Vigo Photonics, Ekspla and, again, Opticoelectron. Sindlhauser, Antraco, Garnet and InPACT fall within materials, chemicals or specialist engineering. Lafert and IHC Merwede manufacture industrial motors and marine systems respectively. Wroclaw University of Science and Technology is the only academic institution.

The individual entities and their most plausible areas of exposure are as follows.

1. Lafert S.p.A. — Italy

Lafert manufactures high-efficiency industrial motors, including IE4 and IE5 permanent-magnet motors, and is part of the Sumitomo group. Its clearest potential exposure is to sintered neodymium-iron-boron magnets and the neodymium, praseodymium, dysprosium and terbium used in high-performance magnet systems. China remains the leading supplier across much of this value chain. Electrical steel may be another area of exposure, although the extent of Lafert’s reliance on Chinese material cannot be established from public information alone.

2. Garnet S.r.l. — Italy

Garnet is an industrial-engineering and specialist-machinery supplier based in Concorezzo. Compared with several other companies on the list, its direct exposure to Chinese controlled raw materials is less obvious. The more likely risk is indirect, through precision components, electronic subassemblies and customer requirements concerning the origin of controlled content.

3. Sindlhauser Materials GmbH — Germany

Sindlhauser supplies sputtering targets, lanthanum hexaboride ceramics and cathodes, bonded rare-earth metals, and tungsten, tantalum and molybdenum components used in coating and vacuum applications. This is one of the clearer supply-chain targets. The company’s product range draws directly on materials in which Chinese processors have a large or dominant market position, including LaB6 feedstock, rare-earth metals and refractory metals.

4. Rheinmetall AG — Germany

Rheinmetall is the largest company on the list and one of Europe’s principal defence contractors, with businesses spanning ammunition, armoured vehicles, air defence and military electronics. Potentially affected inputs include antimony used in ammunition, tungsten used in penetrators and armour, rare-earth materials, magnesium, and gallium- or germanium-containing electronic components. China accounted for a substantial majority of refined antimony production in 2024–2025, although Rheinmetall’s stockpiling and diversification efforts since 2024 may reduce its immediate vulnerability.

5. Antraco Chemie-Handelsgesellschaft mbH — Germany

Antraco Chemie is a Duisburg-based trader in specialty and fine chemicals. Its exposure is likely to arise from direct sourcing relationships with Chinese chemical producers and from its role as an intermediary. The prohibition on overseas transfers of Chinese-origin controlled items is particularly relevant to a trading company, because it can apply even where the goods do not move directly from China to the listed end user.

6. InPACT S.A. — France

InPACT, based in Saint-Marcel in Savoie, specialises in metallurgy and precision-machined components. Its possible exposure lies in alloying materials such as tungsten, molybdenum and certain rare-earth products, as well as specialist steels containing controlled inputs. Public information does not reveal how much of this material the company sources directly or indirectly from China.

7. III-V Lab — France

III-V Lab is a joint research laboratory established by Thales, Nokia Bell Labs and CEA-Leti. It works on gallium arsenide, indium phosphide and gallium nitride epitaxy and related compound-semiconductor components. The laboratory therefore has relatively direct exposure to gallium, indium, germanium and high-purity precursors. Gallium is the most obvious pressure point: China supplies more than 90 per cent of global primary production and has required export licences for gallium-related items since 2023.

8. Cavok UAS — France

Cavok UAS develops professional and defence-oriented unmanned aircraft, including first-person-view systems, from its base in Sainte-Menehould. Its exposure is likely to be concentrated in motors, electronic speed controllers, cameras, infrared or optical modules and other subassemblies that may contain Chinese components. Some of these parts can be replaced, but redesign and requalification are rarely immediate, especially for equipment supplied to military customers.

9. Vigo Photonics S.A. — Poland

Vigo Photonics produces mercury-cadmium-telluride and indium-arsenide-antimonide infrared detectors, epitaxial wafers and infrared modules. Its manufacturing processes depend on specialised semiconductor feedstocks, including gallium, indium, germanium, cadmium and tellurium. Not all of these materials are covered to the same extent by current Chinese controls, but the company is clearly situated in a sector where Chinese licensing decisions can affect availability, lead times and prices.

10. Wroclaw University of Science and Technology — Poland

The university conducts research in materials science, photonics, semiconductors and cybersecurity. Its direct consumption of strategic raw materials is unlikely to be comparable with that of an industrial manufacturer. The more significant consequences may concern research-grade reagents, test wafers, lasers and semiconductor components, as well as access to Chinese partners and joint programmes. The listing may also discourage Chinese institutions from entering or continuing collaborations with the university.

11. IHC Merwede Holding B.V. (Royal IHC) — the Netherlands

Royal IHC builds dredgers, offshore installation vessels and other specialised maritime equipment. Relevant areas of exposure may include permanent-magnet systems used in propulsion and dredging drives, electrical steel and components associated with deep-sea mining equipment. The precise Chinese-origin content of individual vessels will vary by programme and supplier.

12. TATRA TRUCKS a.s. — Czech Republic

TATRA manufactures military and heavy-duty truck chassis, including its Phoenix and Force ranges, and supplies customers in the Czech Republic, Belgium, the Netherlands and Ukraine. Its potential exposure includes tungsten and molybdenum used in high-strength applications, permanent magnets in drive and control systems, and Chinese-origin content in infrared or optronic equipment supplied by third parties.

13. Opticoelectron Group — Bulgaria

Opticoelectron produces optical and optoelectronic equipment, including night-vision devices, thermal sights and gunsights. Germanium optics, infrared-transparent crystals, gallium arsenide and antimony compounds are all relevant to these product lines. China’s position in refined germanium and antimony makes the company a comparatively direct target of material-based export pressure.

14. Ekspla UAB — Lithuania

Ekspla manufactures femtosecond, picosecond and nanosecond lasers, as well as optical parametric chirped-pulse amplification and optical parametric oscillator systems. These products use nonlinear optical crystals such as BBO, LBO, KTP and lithium niobate. Chinese growers are among the leading global suppliers of these crystals, although alternative producers exist in Europe, Japan and elsewhere.

China’s Export Control List is narrower than the Unreliable Entity List. Inclusion does not, by itself, prevent a company from operating in China, freeze its assets or automatically extend to every subsidiary. Its main effect is on access to controlled Chinese-origin inputs.

The first layer is the direct export ban. Chinese suppliers must suspend exports of controlled dual-use items to the 14 entities, and transactions already under way were required to stop from 24 July. Companies such as Sindlhauser, Antraco, Vigo Photonics, III-V Lab, Ekspla and Opticoelectron are likely to feel this effect most directly. Their options include applying to MOFCOM for an exceptional licence or shifting procurement to suppliers in Japan, South Korea, the EU or the United States. Whether those alternatives are commercially viable will differ sharply by material and product grade.

The second layer is the restriction on overseas re-transfers. Distributors and resellers outside China may not supply the listed entities with Chinese-origin dual-use items. This potentially affects all 14 entities, including firms that do not buy directly from Chinese exporters. Rheinmetall, TATRA, Royal IHC and Lafert, for example, may receive controlled Chinese-origin content through European, American or Japanese intermediaries. Suppliers will therefore need stronger end-user declarations, origin audits and contractual restrictions. The compliance burden resembles some aspects of US export-control practice, although it should not be treated as legally equivalent to the US Foreign Direct Product Rules.

The third layer consists of contractual, banking and insurance effects. Customers, lenders and insurers may add the Chinese list to their screening systems even where a particular transaction does not involve a controlled item. Delivery dates and prices may need to be renegotiated. This is especially relevant to Rheinmetall’s NATO and Ukraine-related contracts, TATRA’s military-vehicle orders and Royal IHC’s long-term offshore projects. Companies may respond through public disclosures, hardship or force-majeure provisions, and formal plans to remove Chinese-origin content from sensitive product lines.

The fourth layer is reputational and institutional. Chinese partners may become reluctant to work with Wroclaw University of Science and Technology, III-V Lab or listed companies with joint ventures or research ties in China. The likely responses are fairly conventional: appointing a designated compliance lead, reviewing joint projects, separating controlled and non-controlled work, and engaging MOFCOM on the availability of licences.

The economic effect will not be uniform. Rheinmetall, with more than €10 billion in annual group revenue in 2025, is by far the largest listed company. Royal IHC, Lafert, TATRA and the industrial groups behind III-V Lab operate on a different but still substantial scale. Vigo Photonics, Ekspla, Opticoelectron and Sindlhauser are smaller specialist suppliers, broadly in the tens to low hundreds of millions of euros in revenue. For these firms, the loss of a particular Chinese input may be more painful as a share of turnover, even if the absolute economic value is smaller. The university’s exposure is primarily institutional rather than financial.

Alternative supplies of gallium, germanium, antimony, rare-earth materials and nonlinear crystals do exist. The question is less whether substitution is technically possible than how long it takes, whether the replacement meets the required specification, and at what cost. For some high-purity materials or specialised optical crystals, qualification can take years and unit costs can increase several-fold. Larger defence companies may have inventories and long-term procurement arrangements. Smaller photonics and materials firms are less likely to have the same buffers.

The overseas re-transfer clause deserves particular attention. Its wording is broad and resembles language used in China’s 2024–2026 actions against US companies. In practical terms, an EU distributor handling controlled Chinese-origin goods will need to screen the 14 entities as end users and consignees. The compliance risk therefore rests not only with the original Chinese exporter but also with intermediaries and the final recipient.

What to watch next？Three wider consequences stand out.

First, China has demonstrated that it can respond to an EU listing within a day and on a list-for-list basis. Future EU sanctions packages that include Chinese entities are therefore more likely to trigger a targeted Chinese control-list response rather than a prolonged diplomatic protest. This changes the calculation for member states that may previously have regarded the inclusion of Chinese companies as a relatively low-cost measure.

Second, the listing of a public university crosses a line that MOFCOM had largely avoided in its earlier measures against European entities. It puts national research ministries and universities on notice that Chinese countermeasures can reach laboratory supply chains and institutional partnerships. Researchers working in materials science, photonics and semiconductors will need to pay closer attention to dual-use classification in future collaborations with Chinese counterparts.

Third, the announcement leaves important implementation questions unanswered. MOFCOM has not yet explained what will qualify as an “exceptional case”, whether spare parts and after-sales services for previously installed Chinese-origin equipment are covered, or how transactions routed through Hong Kong will be treated. Its first licensing decisions will matter more than the wording of the announcement in determining the practical severity of the measure.

Appendix A. Full English translation of MOFCOM Announcement No. 30 of 2026

Translation provided for reader convenience; not an official version.

Announcement No. 30 of 2026 of the Ministry of Commerce of the People's Republic of China: Announcing the Inclusion of 14 EU Entities on the Export Control Control List

Issuing unit: Bureau of Security and Control

Document number: MOFCOM Announcement No. 30 of 2026

Date of issue: 24 July 2026

Pursuant to the relevant provisions of the Export Control Law of the People's Republic of China, the Regulations of the People's Republic of China on the Export Control of Dual-Use Items and other laws and regulations, in order to safeguard national security and interests and to fulfil international obligations such as non-proliferation, it is decided to include 14 EU entities, including the Lafert Group, on the Export Control Control List (see the Annex), and to adopt the following measures:

I. Exporting operators are prohibited from exporting dual-use items to the 14 entities listed above; foreign organisations and individuals are prohibited from transferring or providing dual-use items originating in the People's Republic of China to the 14 entities listed above; any related activities currently ongoing shall be terminated immediately.

II. Where an export is genuinely required in exceptional cases, the exporting operator shall submit an application to the Ministry of Commerce.

This announcement takes effect on the date of its publication.

Annex: Export Control Control List (24 July 2026)

Ministry of Commerce

24 July 2026

Annex — Export Control Control List (24 July 2026)

1. Lafert Group (Lafert S.p.A.)

Address: Via J.F. Kennedy 43, San Donà di Piave, Venice, Italy

Postal code: 30027

2. Garnet Company (Garnet S.r.l.)

Address: Via De Gasperi 31, Concorezzo (MB), Italy

Postal code: 20863

3. Sindlhauser Materials Ltd. (Sindlhauser Materials GmbH)

Address: Daimlerstraße 68, Kempten, Germany

Postal code: 87437

4. Rheinmetall Company (Rheinmetall AG)

Address: Rheinmetall Platz 1, Düsseldorf, Germany

Postal code: 40476

5. Antraco Chemical Trading Ltd. (Antraco Chemie-Handelsgesellschaft mbH)

Address: Düsseldorfer Landstraße 17, Duisburg, Germany

Postal code: 47279

6. InPACT Company (InPACT S.A.)

Address: 265 Rue de la Volta, Saint-Marcel, France

Postal code: 73600

7. III-V Lab (III-V LAB)

Address: 1 avenue Augustin Fresnel, Palaiseau, France

Postal code: F-91767

8. Cavok UAS Company (Cavok UAS)

Address: Allée des Cuirassiers, Sainte-Menehould, France

Postal code: 51800

9. Vigo Photonics Company (Vigo Photonics S.A.)

Address: 129/133 Poznanska St., Ozarow Mazowiecki, Poland

Postal code: 05-850

10. Wrocław University of Science and Technology (Politechnika Wrocławska)

Address: Wybrzeże Stanisława Wyspiańskiego 27, Wrocław, Poland

Postal code: 50-370

11. IHC Company (IHC Merwede Holding B.V.)

Address: Smitweg 6, Kinderdijk, The Netherlands

Postal code: 2961

12. TATRA Trucks Company (TATRA TRUCKS a.s.)

Address: Areal Tatry 1450/1, Kopřivnice, Czech Republic

Postal code: 74221

13. Opticoelectron Group (Opticoelectron Group)

Address: Industrial Park Opticoelectron, Panagyurishte, Bulgaria

Postal code: 4500

14. Ekspla Company (Ekspla UAB)

Address: Savanoriu Ave 237, Vilnius, Lithuania

Postal code: LT-02300

Appendix B. Full English translation of the MOFCOM spokesperson Q&A

Translation provided for reader convenience; not an official version.

Spokesperson of the Ministry of Commerce answers press questions on the inclusion of 14 EU entities on the Export Control Control List

Question. On 24 July 2026, the Ministry of Commerce made public the Control List. What are the considerations behind this?

Answer. On the evening of 23 July, Beijing time, the EU side formally released its 21st package of sanctions against Russia, which imposed listing sanctions on 14 mainland Chinese and Hong Kong enterprises. In order to safeguard national security and interests and to fulfil international obligations such as non-proliferation, and in response to the aforementioned egregious behaviour of the EU side, in accordance with the relevant provisions of the Export Control Law of the People's Republic of China, the Regulations of the People's Republic of China on the Export Control of Dual-Use Items and other laws and regulations, China has decided to include 14 EU entities, such as the Lafert Group, on the Export Control Control List, prohibiting exporting operators from exporting dual-use items to the aforementioned entities and prohibiting foreign organisations and individuals from transferring or providing dual-use items originating in the People's Republic of China to the aforementioned entities.

Key takeaways:

• Vision-driven: No KPIs, no written vision, even "no organization"—it runs on goodwill toward the world and an obsession with AGI. "We're just a bunch of very ordinary people."

• Restraint as strategy: API pricing aims only to recover hardware costs in ten months. No fighting for consumer traffic, no chasing enterprise fads. "The more restrained you are, the more likely you'll pull it off."

• Open source is the point: Even the strongest model will be open-sourced. Not worried about others competing. "I only worry they won't be able to deploy it."

• AGI roadmap: CoT → Agents → continual learning → "singularity" → embodied intelligence. "After continual learning, it can develop the next version itself."

• Compute reality: About 20,000 "H-equivalent" GPUs; 12–18 months behind the U.S.; chasing with one‑twentieth the compute. For domestic chips: "The ecosystem is fine; capacity is the bottleneck."

• Endgame view: China may match foreign models this year, but "that still isn't AGI." "There are too many base-model companies in China; it will converge."

Vision, motivation, and culture

*"We don't have an organization; we're driven by a vision."*

1. When we started the company, our original intent wasn't "how much money will I make," or "go to the capital markets," or "IPO," or anything like that. That wasn't the intent. The first few dozen people never thought that way. If they did, they wouldn't have joined.

2. Overall, we're doing this with a lot of goodwill toward the world, and we think it's useful to humanity. It's something beyond money.

3. About twenty years ago, in management, the person I admired most was Jack Welch, former CEO of GE. Looking back now, most of what he said may no longer be right, but he got one thing right: the most important thing for a company is its vision. Managing a big company isn't about rules and procedures—it's about vision. What is vision? Vision isn't a slogan on the wall. Vision is how you do things, not what you say.

4. In fact, we don't really have an organization—we're driven by a vision, organized by a vision. That vision isn't even written down. We've never written anything.

5. We're not especially capable. We don't have more money than others, and it's not that our people are better than everyone else's. We're just a bunch of very ordinary people.

6. Our company is built on consensus. It's not that I decide everything. I seek consensus. My authority inside the company is built on consensus.

7. Our management has two lines: one top-down and one bottom-up. Bottom-up means everyone decides what they want to do and just does it. No one manages them. No KPIs. In general, we hope "formal work" doesn't take more than half of an employee's time. The other half is unallocated—they can do whatever they want.

8. We generally don't work overtime. First, research needs a relaxed environment; if you push too hard, you can't do research. Second, we're very focused, which means we do very few things.

9. CoT doesn't consume resources. For research, it doesn't burn GPUs—you only need a small number of cards. What you need is ideas. The barrier is low; anyone can try it. But who can actually get something out of it—I don't know if that's talent or what.

Open-source strategy

*"Even our strongest model will be open source, because I don't see what's good about closed source."*

1. Why do we insist on open source? Because the vision itself requires open source. Without that vision, you can't organize people.

2. Zhipu also open-sources, but their open source isn't the same as ours. Their open source has a "being chased" feeling—they don't see it as the point. For us, it is the point.

3. AI is big enough that in the end it may take up 10% of human society's GDP. One person can't monopolize it. You have to share it with others, or you won't survive.

4. I think we will open-source, and our strongest model will probably also be open-sourced, because I don't see what's good about closed source.

5. Even if you open-source the model and tell people everything, the barrier is still very high. It's still very hard for others to actually use it.

6. If we only make, say, a 6× profit, open source doesn't really hurt. But if you want a 100× profit, then open source does affect that.

7. We were open-source from the start. We won't treat you differently just because you're a competitor—that's part of open source. I'm not worried about others deploying our model to compete with us. I want them to be able to deploy it. I only worry they can't—some detail is wrong, results get worse, or costs get high.

8. Are the open-source models we provide the same as what we deploy ourselves? Yes, they're the same. We won't open-source a weaker model and then use a better one internally. Last year, basically everything on the consumer side was open source, and we didn't see a conflict with our consumer service.

Restraint and pricing

*"Restraint is a strategy. Ten months to recoup costs is a reasonable profit."*

1. If we want to make AI work in our hands, the first thing, I think, is restraint. You can't think "a huge share of human GDP should be mine." The more you think like that, the less you'll succeed.

2. For me, restraint is a strategy. Sometimes you give up something to gain other things.

3. For our API pricing, what we consider a reasonable profit is: buy a batch of equipment from the market, and recoup the cost in ten months. I think that's reasonable. It's not profit-maximizing. If you wanted to maximize profit, you'd price higher. In this price range, demand isn't elastic. Even if I raise the price by 50%, token usage doesn't change much.

4. Let me tell a story. For our DCP (Decoder Context Parallelism) model, at first we worried demand would be too high, so we set the price relatively high. People on the team weren't very happy. Later I lowered it—down to one quarter—and everyone was happy. When we cut the price, people in the company group chat were cheering.

5. Someone just commented that "ten months to break even is too high." True, there's still room to lower prices. There's room for optimization on the model side too, so there's still a lot of room overall.

6. At that cost level—ten months to break even—we can do it. Other companies can't. Alibaba or Tencent, without our optimizations, their costs should be several times higher.

7. If I cut prices further, demand won't increase, or will increase only slightly. At this price, everyone can afford it and is satisfied. Cutting prices won't bring the company more revenue, and it won't add more value to society. Even if it's cheaper, it doesn't increase "happiness" by much.

8. The lower the cost, the bigger the model I can train, and the more I can afford to take on larger models. With limited compute, if my compute efficiency is higher, I can take on larger models.

9. A commercial company has no incentive to pursue model efficiency. If efficiency improves… costs drop… then what do you earn? For us, it's part of the vision. Our teammates are ordinary people; they know using this costs money. They have empathy: others have to pay to use it, so if it's cheaper, it's easier to accept.

Commercialization

*"Both C-end and B-end are side products on our road to AGI."*

1. We do AGI to do AGI. We just happen to produce something we can commercialize, so we use it. That's different from other companies. Other companies build the model to serve consumer users or enterprise users. For us, both consumer and enterprise are side products on the road to AGI.

2. When we suddenly got popular around last year's Spring Festival, that wasn't in our script. We never expected it. We just wanted to make the technology good.

3. When users suddenly surged last year, we didn't try to keep them for monetization, or fight for those commercial interests. We didn't fight for users, we didn't try to make money—but we worked hard to serve them well.

4. This year it's quite possible our API/AI revenue could reach a few hundred million dollars in ARR, if demand continues expanding and if we can buy more GPUs. If AI revenue can reach a billion dollars, our cash flow can basically turn positive.

5. Over the past three years, at any point, if you came to me to talk about a commercialization roadmap or product lines, it would be a waste of time, because you can't predict the future.

6. If this year we can have a few hundred million dollars of enterprise revenue, and next year enterprise revenue continues, and demand keeps growing, we won't be far from net profit. In the worst case, just selling APIs can support a listed company. If there's no new technical progress later and our technology "freezes" here, then we'll fully focus on selling APIs and doing the services well—that's enough.

AGI tech roadmap

*"AI isn't lacking taste or intuition. It's lacking continual learning."*

1. You can think of AI progress as climbing steps. Last year's step was CoT. This year's step is Agents. Why steps? Because each next step builds on the previous one. Agents rely on CoT, and CoT relies on the previous step—the language model.

2. After Agents, we think the next problem to solve is continual learning: how to let the model keep learning continuously, rather than just giving it one strong training run. It should be able to learn over a long period like humans do.

3. After continual learning, we may reach a "singularity." It's when a model that can keep learning can already do everything humans can do. It can develop its own versions, do research by itself, and build the next version—build even more advanced AI models.

4. This "singularity" isn't really a point; it's also a gradual process. It may be a long, gradual change—not a sudden jump.

5. After that, I think you get embodied intelligence. Then it enters the real world: it can do housework, take care of the elderly, and so on.

6. We think this roadmap is the easiest. We can follow it without overtime. If the roadmap were reversed—for example, if you try to do embodied intelligence first—you'd suffer. It's hard labor.

7. AI isn't lacking taste or intuition. Its taste and intuition are fine. If you ask it to write an article, I think its taste and intuition are basically fine. What it's lacking is continual learning.

8. Everyone in the world is researching this. From an investor's perspective, what they see most is Agents. But for researchers like us, what we see more is learning, and how to solve the learning problem.

9. Internally we care about this narrative: for the next version of our model, we hope it can help our own development. It can improve DeepSeek's efficiency. Our model's first job is to raise DeepSeek's own work efficiency. First it has to be useful to us. That's the fastest path to AGI.

Compute, chips, and the China–U.S. gap

*"We're behind the U.S. mainly on resources. On people, almost no gap."*

1. Right now we have about 20,000 H-equivalent compute cards, and most of those arrived recently—within the past one or two months.

2. Our gap with the U.S. is mainly resources. On people, almost no gap, because it's basically the same group of people. Talent isn't the bottleneck; resources are the biggest bottleneck. The talent gap is essentially also a compute gap.

3. For the largest models today, we can't afford to train them. The biggest models have about 800B active parameters; domestically, we're still at tens of billions. If I want to train a model as large as the leading ones, I'd need 50,000 GB300s or Huawei 950s—200,000 cards. That's just training, not even research.

4. Within what we can afford, more cards is always better. At a reasonable price, we buy as many as we can. Ideally, if we can spend the money within six months, that's best. Turning money into Nvidia cards is definitely better than leaving it in the bank.

5. Our gap with the U.S. may be 12 months, maybe 12–18 months, or maybe 6–12 months. Simply put: about two years behind, but doing it with one-twentieth of the U.S. compute. In the future we want to rewrite that narrative: use a fraction of the compute, but shorten the time to 6 months, 3 months—maybe even surpass them in some areas.

6. Nvidia's CUDA moat is being eroded quickly. One reason is that AI can now write code, so I can use AI to build the ecosystem. Another reason is new technologies—for example TileLang (an open-source high-performance AI operator programming language developed by Peking University's School of Computer Science). Using a higher-level language to write CUDA operators, you can quickly rewrite Nvidia's whole ecosystem.

7. There's a historic opportunity for domestic AI chips to replace imports. We believe within the next year we'll see something validated: the ecosystem for domestic chips has no problem. People used to think there were problems—couldn't use them, not usable—but within a year, I think we can change that perception.

8. Our goal in buying Huawei 950 is still to help Huawei improve the ecosystem. Huawei 950 "super nodes" can substitute Nvidia GB200/GB300 in performance and price. The only tradeoff is: four Huawei cards equal one Nvidia card, and you're two years behind in time. Huawei 950 super nodes will ship in Q3 or Q4 this year; Nvidia GB200 shipped in Q3 two years ago.

9. You can depreciate Nvidia cards over five years. Huawei cards, at most three years. Huawei 950 is fine to use this year; next year should still be OK. After that, it may really become too power-hungry.

What DeepSeek won't do

*"Video generation, 3D, world models—these don't set the upper bound of intelligence."*

1. AI is a broad field. There are many things we think aren't on the main line—for example 3D and video generation. I don't think they're strongly related to the main line of intelligence, so we won't do them.

2. When video generation first came out, it was very hot—like you had to do it, or you weren't an AI company. But you can see what happened: after Sora, everyone did it—big companies, small companies. Then small companies cut it. It doesn't affect the ceiling of intelligence. Commercially, it's a good business, but it's not about intelligence. We won't do something just because it's a good business.

3. World models, I think, also don't have much to do with the upper bound of intelligence right now, so we won't do them. In our view, world models and intelligence aren't the most important thing at this stage.

4. Multimodal matters a lot for products, and for consumer products. But for the ceiling of intelligence, it's a component, not the main line itself. We will definitely do multimodal, and we're doing it. V4 and later versions will support native multimodal.

Data labeling

1. The cost of data labeling in the U.S. isn't really different from China. China doesn't have a cost advantage in labeling, especially for high-end labeling. That makes it hard for us to invest in labeling data the way the U.S. does. This path is hard in China because labeling is just too expensive.

2. Right now, half of the most important people in the company are labeling data.

3. The bottleneck isn't that we can't quickly hire more people. It's not about money, and not about cards either. But we are expanding fast. So we think within a year, it's reasonable to expect that the high-quality data problem in China can be handled much better.

Competition and endgame

*"There are too many base-model companies in China. It will converge."*

1. Our biggest core interest is keeping the team stable. That's our biggest core interest—maybe the only one. As long as I can keep the team stable, I'll be able to do AGI. Money isn't the problem, resources aren't the problem; everything else is easy to get.

2. We've always been very restrained. We don't want to become an opponent of any big internet company or small company. I'd rather help them do this.

3. Open source, goodwill, helping others—none of that caused us to lose anything. No impact at all. If anything, it may be a plus.

4. There are a bit too many model companies in China right now—still too many. The U.S. maybe has three. China has too many doing base models. Everyone is doing the same thing. Resources get spread thin, and each company gets less. I think it will definitely converge.

5. If my goal is to take 5% of global GDP from AI, I'll be defeated by someone willing to take only 1%. Because that person says: I do it well, but I only need 1% of global GDP. Then if another person comes along and says: I only need 0.1%, they beat the previous one. People who take more get beaten by people who take less. OpenAI at the beginning thought it could monopolize the world, but in reality it will face many challengers.

6. In AI, in China, in one or two years we can get to roughly the same level as abroad, or maybe even substitute foreign models this year. In the current paradigm, it's not that hard. But it's still not AGI.

7. Right now, returns from model scaling are still very obvious. We haven't had a chance to hit the "scaling wall." It's still far away. When Silicon Valley says scaling is over, that's for Silicon Valley. For China, we're still far from that—we haven't scaled to that level at all.

Source: https://mp.weixin.qq.com/s/HxvyxDGXusNRMzes3i-U7g

Some comment：

The “mutual trust, interconnection, and interoperability” of intelligent agents, in plain terms, is about whether AI agents built by different companies can recognize each other, communicate with each other, and work together.

The current reality is that OpenAI’s agents, Google’s agents, Microsoft’s agents, and the various Chinese agents all use their own formats and their own interfaces — they largely don’t speak the same language. If you try to get one agent to call another company’s agent, it either fails entirely or requires writing substantial amounts of glue code. The deeper issue is trust: why should one agent believe another agent’s identity, or accept the instructions it passes along? There is no unified answer to this yet.

In the era of single-machine, single-platform software, this wasn’t a major problem. But the entire value proposition of agents is that they can “autonomously complete complex tasks” — and complex tasks almost always require calling multiple tools and crossing multiple systems. If the agent ecosystem remains a collection of isolated islands, the efficiency ceiling of the entire industry gets permanently capped.

Why does agent interoperability matter? The most direct reason is that whoever sets the standard controls who gets to enter the ecosystem. When Wi-Fi, HTTP, and TCP/IP were established, everyone played by those rules and network effects could take off. Agent interoperability standards follow the same logic — once a standard solidifies, the cost of retrofitting systems for newcomers becomes extremely high, and whoever staked out the position first builds a long-term structural advantage.

For developing countries, there is an additional dimension: if interoperability standards are quietly set by a handful of large platforms, smaller countries and businesses have no choice but to passively plug into someone else’s ecosystem, while data, traffic, and profits concentrate at the top. This is what the initiative refers to as the “intelligence divide.”

On the US side, the approach is primarily industry-driven, with no unified national push. A few substantive developments:

• MCP (Model Context Protocol): An open protocol proposed by Anthropic in late 2024 that defines how AI models call external tools and data sources. It has already been adopted by a large developer community, with OpenAI and Google subsequently adding support. It is currently the most influential candidate for a de facto standard.

• A2A (Agent-to-Agent): Proposed by Google in 2025, specifically targeting communication protocols between multiple agents. Its scope overlaps with MCP; the two currently coexist and have not yet been merged.

• NIST: The US National Institute of Standards and Technology is working on an AI safety and interoperability framework, but the pace is slow. It is primarily a reference for government procurement rather than a mandatory standard.

Overall, the US approach is to “let the market sort it out” — whichever open-source protocol gets used the most becomes the de facto standard. Anthropic’s MCP currently has the strongest momentum, but the outcome is not yet settled.

The initiative China published today represents the first time China has systematically articulated its position on this issue. Released at a high-profile international venue like WAIC, it is directed at signatories and audiences among the Global South. The “agent identity identification” and “full lifecycle safety governance framework” proposed in the initiative are designed to complement China’s existing domestic regulatory framework — establish the standard at home first, then extend it outward.

From an industry perspective, Alibaba, Baidu, Huawei, and Tencent are all building their own agent frameworks (Bailian, ERNIE Agent, Pangu, etc.), and the interoperability problem exists among them as well. The initiative carries a domestic market integration intent, not just an outward-facing posture.

Global Cooperation Initiative on Agent Mutual Trust, Interconnection, and Interoperability

As one of the most transformative technological forms of the AI era, intelligent agents — and the degree to which they can mutually trust, interconnect, and interoperate — will profoundly shape the integration of the global digital economy and the future of human society. To unlock the potential of agents in enabling sustainable development and to prevent the emergence of an intelligence divide, on July 17, the Cyberspace Administration of China, together with relevant parties, put forward this Global Cooperation Initiative on Agent Mutual Trust, Interconnection, and Interoperability at the main forum of the 2026 World Artificial Intelligence Conference and High-Level Meeting on Global AI Governance. We hope to build broad consensus and work with global partners to foster an open, trustworthy, safe, and inclusive agent ecosystem.

I. Strengthen R&D to Drive Collaborative Innovation

We call on all countries to increase investment in foundational and shared technologies — including agent identity identification and collaborative mutual recognition — to drive the evolution of multi-agent systems from “controlled interaction” toward “autonomous collaboration,” and to build efficient, scalable, and low-latency cross-platform interoperability capabilities, laying a solid technological foundation for global agent interconnection.

II. Promote Standards Mutual Recognition to Eliminate Ecosystem Barriers

We support the development of open, non-discriminatory, and transparent international interoperability standards for agents on the basis of broad participation by all countries, advancing compatibility and mutual recognition at the interface, interconnection protocol, semantic, and process layers. We advocate establishing open standards validation and iteration mechanisms to avoid technical barriers and ecosystem fragmentation, enabling different agents to seamlessly join global collaborative networks.

III. Co-Build Infrastructure to Ensure Accessible Interconnection

We encourage all parties to jointly study architectures for agent interconnection and collaboration, to co-build open and shared smart internet infrastructure serving efficient and trustworthy multi-agent collaboration, and to ensure that such infrastructure has strong compatibility and scalability so as to lower the threshold for integration and adaptation.

IV. Jointly Advance Deployment to Accelerate Scenario Demonstrations

We advocate applying the outcomes of agent mutual trust, interconnection, and interoperability broadly to scientific research, industrial development, consumption stimulation, public welfare, and social governance, with priority given to addressing global challenges such as climate change, public health, educational equity, and poverty reduction — ensuring that the benefits accrue to humanity as a whole.

V. Promote Industrial Coordination to Cultivate a Thriving Global Ecosystem

We welcome broad participation from developed and developing countries, large enterprises and small and medium-sized businesses, research institutions and open-source communities in building agent mutual trust, interconnection, and interoperability. We encourage the wide dissemination of knowledge, technology, and best practices globally to form a global agent industrial chain.

VI. Collectively Uphold Safety Baselines and Adhere to Ethical Principles

We call on all parties to maintain a fundamental commitment to responsible innovation, to establish a safety governance framework covering the full lifecycle of agents, and to embed safety and ethical considerations into technology R&D, system deployment, and operation and maintenance. We emphasize the profound implications of agents for cybersecurity, and call for the responsible release and use of agents, ensuring that agent mutual trust, interconnection, and interoperability never crosses the red lines of safety and human dignity.

VII. Facilitate Data Flows While Strengthening Universal Protection

We call on all parties to respect differences in data security and personal information protection regimes across countries and regions, and to advocate for establishing a cross-border data cooperation framework based on openness, inclusiveness, security, cooperation, and non-discrimination, enabling data to flow safely, compliantly, and in an orderly manner throughout agent mutual trust, interconnection, and interoperability processes.

VIII. Enhance Resource Sharing to Bridge the Intelligence Divide

We encourage the provision to developing countries of infrastructure development support, open-source tools, capacity training, and financial and technical assistance needed for agent mutual trust, interconnection, and interoperability. We encourage the adoption of localized adaptation and capacity co-building approaches, and promote the formation of a new “South-South plus North-South” digital cooperation model so that the dividends of the intelligent economy are shared inclusively.

IX. Foster an Inclusive Culture and Respect Diverse Governance

We advocate cross-cultural, cross-linguistic, and interdisciplinary norms for agent collaboration, respect different countries’ governance models and value choices, and oppose unilateral imposition. We encourage broad participation by diverse stakeholders in rule discussions, to foster a global agent collaboration culture that balances efficiency with fairness, and flexibility with order.

X. Enhance Digital Literacy to Ensure Everyone Benefits

We will work together — through capacity building, technical assistance, and the opening of educational resources — to help minors, the elderly, persons with disabilities, and low-skill groups improve their AI literacy, ensuring that the outcomes of agent mutual trust, interconnection, and interoperability benefit all types of actors.

We uphold the principles of extensive consultation, joint contribution, and shared benefits, and sincerely invite all parties who share this vision to join us — with an open and inclusive mindset and prudent, practical action — in jointly advancing the development of the agent ecosystem and contributing wisdom and strength to the building of a community with a shared future in cyberspace.

AI Cooperation Development Action Plan

National Development and Reform Commission

July 17, 2026, 12:43

Beijing

1. High-Quality Data Supply Initiative

Promote cross-border data flows, and in selected fields build and operate trusted cross-border data spaces to enable efficient, convenient, and secure cross-border movement of data. Work together to develop high-quality corpora and industry-specific high-quality datasets, and advance the co-building and sharing of multilingual corpora, so as to lay a solid foundation for global AI innovation.

2. Inclusive Intelligent Computing Initiative

Promote connectivity among intelligent computing infrastructure. Provide inclusive intelligent computing services for developing countries. Jointly build infrastructure powered by green energy, advance the green and low-carbon development of intelligent computing facilities, and provide computing support for the sustainable development of AI worldwide.

3. Shared Open-Source Ecosystem Initiative

Encourage the joint building of international AI open-source communities, carry out international exchange and cooperation among open-source communities, and promote the sharing of general-purpose foundation models, core algorithms, and tool components. Work together to formulate open-source compliance systems and safety guidelines. Support countries in pursuing localized innovation based on open-source models, and build a global AI open-source ecosystem that is open and shared, secure and orderly, and collaboratively governed.

4. Deep AI Enablement Initiative

Deepen cooperation under the “AI+” framework, build cross-border industrial cooperation platforms, support digital and intelligent capacity-building for developing countries, promote the standardized application and innovative development of AI agents, and advance AI-enabled applications in science, manufacturing, healthcare, education, agriculture, and governance. Use AI to promote economic development, improve social governance, and enhance public well-being.

5. Joint Digital and AI Talent Development Initiative

Jointly build mechanisms for the collaborative training of top-tier digital and AI talent, and promote international cooperation among universities, research institutions, and enterprises. Work together to formulate occupational standards and skills certification systems, and provide diversified, interdisciplinary talent support for global AI development. Raise public AI literacy while paying attention to and safeguarding the rights and interests of vulnerable groups. Jointly address the impact of rapid AI development on employment.

6. Joint Rules and Standards Initiative

Jointly build a system of AI standards and norms, and promote the formulation and revision of international standards. Work together to establish mechanisms for standards alignment and coordination, and improve the inclusiveness and interoperability of standards systems. Strengthen coordination and alignment in AI development strategies, governance rules, and technical standards.

7. Collaborative Safety and Governance Initiative

Jointly build mechanisms for AI safety governance, strengthen the sharing of cybersecurity threat information and cooperation in emergency response, and guard against the misuse and abuse of AI technologies. Study ways to improve the explainability, transparency, and safety of AI, and strengthen AI data governance, in order to foster a sound environment for global AI development. Advocate the establishment of open platforms for sharing best practices and advancing international cooperation on AI safety governance worldwide.

8. AI for Good Initiative

Uphold AI ethics principles of openness and transparency, privacy and security protection, and controllability and trustworthiness, and jointly build a system of AI ethical guidelines. Put into practice the people-centered principle and the concept of AI for good, and promote the building of a more humane intelligent society. Eliminate racism, discrimination, and other forms of algorithmic bias, and safeguard fairness and non-discrimination. Advance international cooperation on the governance of AI research, contribute global public scientific goods to the Global South, and support the United Nations 2030 Sustainable Development Goals.

Joining Hands to Build a Just and Reasonable Global AI Governance System

Distinguished colleagues and guests, ladies and gentlemen, friends,

Seventy years ago, a group of young scholars first put forward the concept of artificial intelligence at the Dartmouth Conference in New Hampshire, the United States. Over the past seventy years, AI scientists and researchers around the world have kept exploring the unknown, moving forward through twists and turns, and making breakthroughs through perseverance. Seventy years later, as we face a new wave of vigorous AI development, it is of great significance that we are gathering here on the banks of the Huangpu River to discuss how to promote the development of global AI in a positive and beneficial direction for humanity. On behalf of the Chinese government and the Chinese people, I extend a warm welcome to all of you.

Looking back at history, the invention of the steam engine ushered in industrial civilization, the spread of electricity lit up modern society, and the birth of the Internet connected the entire world. Every technological revolution has profoundly reshaped the way human beings live and work, driving major leaps in economic and social development.

At present, changes unseen in a century are accelerating across the world, while a new round of technological revolution and industrial transformation is making faster breakthroughs. Global innovation in AI technology has entered an unprecedentedly active period. Intelligent technologies featuring ubiquitous connectivity, human-machine collaboration, cross-sector integration, and co-creation and sharing are releasing enormous energy in combination. They contain tremendous opportunities, but they also present governance challenges. Humanity has no choice but to confront the questions of our time directly: when machines begin to think, how should humans coexist with them? When algorithms take part in decision-making, how can security be ensured? When technology challenges ethics, how can governance keep pace? When gaps continue to widen, how can inclusiveness be achieved? These are questions the international community must think deeply about and answer together.

China believes that all countries should uphold a people-centered philosophy and the principle of developing AI for good, so that artificial intelligence can become an important source of momentum for shared prosperity and common security, and work together to build a just and reasonable global AI governance system. In this regard, I would like to offer four proposals.

First, uphold openness and win-win cooperation to drive innovation and development. Artificial intelligence is a new engine of global economic growth and an accelerator for the transformation of old and new growth drivers. It is moving from the "digital world" into the "physical world." We must seize this rare historic opportunity, encourage open source, openness, cooperation, and sharing, and comprehensively promote AI technological innovation, industrial development, and scenario-based applications. We should advance in coordination the upgrading of traditional industries, the growth of emerging industries, and forward-looking planning for future industries, so that AI can empower all sectors of the economy and society.

Second, strengthen risk awareness to ensure safety and controllability. Artificial intelligence should become a trustworthy tool for humanity. We must attach great importance to the various endogenous and derivative risks brought about by AI, and work to build systems of laws and regulations, technical monitoring, risk early warning, and emergency response. We must reinforce the safety baseline, prevent misuse and abuse, and ensure that artificial intelligence always remains under human control. At the same time, we should jointly oppose the practice of overstretching the concept of national security in the field of AI or placing one country's own security above the security of others.

Third, encourage inclusiveness and mutual learning among civilizations. The development and application of artificial intelligence should not erode or undermine the diversity of world civilizations or the unique cultures of individual countries. We should shape the values of artificial intelligence with the common values of all humanity, make good use of AI technologies to enhance understanding and inclusiveness among different civilizations, promote exchanges and mutual learning among them, and carefully cultivate a flourishing garden of civilizations in which each beauty is appreciated in itself and all beauties are shared together.

Fourth, advocate solidarity and cooperation to improve global governance. Artificial intelligence is the shared crystallization of human wisdom and a valuable asset for all humanity. We should practice true multilateralism, give full play to the important role of the United Nations, strengthen coordination on AI development strategies, governance rules, and technical standards, and form a global governance framework with broad consensus as soon as possible, so that this cutting-edge technology may better benefit human society. We should also carry out extensive international cooperation, help countries of the Global South strengthen capacity building, bridge the digital and intelligence divide, promote sustainable development, and avoid creating new historical injustices in the field of AI.

Ladies and gentlemen, friends,

This year marks the opening year of China's 15th Five-Year Plan period. The 15th Five-Year Plan lays out a blueprint for China's economic and social development over the next five years, and it also provides a list of opportunities for the international community. In recent years, China has actively embraced AI development, adhered to a combination of an efficient market and a well-functioning government, strengthened AI technological innovation, actively advanced the "AI+" initiative, and cultivated a healthy ecosystem in which all kinds of entities coexist and thrive together. The scale of the core industries of the intelligent economy has already exceeded 1 trillion yuan. Various intelligent terminals have entered ordinary households and tangibly improved people's lives. "Intelligent manufacturing in China" has become another striking hallmark of Chinese modernization.

At the same time, China has consistently adhered to the principle of giving equal weight to development and security. It has grasped the trends and laws of AI development, continuously improved related laws and regulations, policy systems, application norms, and ethical guidelines, and ensured that AI is safe, reliable, and controllable, so that this fine steed of artificial intelligence can run both fast and steadily.

As a responsible major country, China has always been committed to providing international public goods in the field of artificial intelligence. Since I proposed the Global AI Governance Initiative, China has successively promoted the adoption by consensus at the UN General Assembly of a resolution on international cooperation in AI capacity building, issued the AI Capacity-Building for All Plan and the "AI+" International Cooperation Initiative, and proposed the establishment of a World AI Cooperation Organization, continuously contributing Chinese solutions.

As the Chinese often say, a single string cannot make music, and a lone tree does not make a forest. The development of artificial intelligence should not be a solo performance by any one country, but a symphony of global cooperation. Through the joint efforts of all parties, the World AI Cooperation Organization has come into being in Shanghai, turning an idea from one year ago into a concrete achievement. This is a major measure by China in response to the calls of the Global South and in working with the international community to actively advance AI development and governance. It will become an important milestone in the history of AI development.

To further support global AI development and advance global AI capacity building, I announce that over the next five years, China will provide 5,000 training places in AI-focused programs for developing countries; establish international cooperation centers for AI applications with ASEAN, the League of Arab States, the African Union, the Community of Latin American and Caribbean States, the Shanghai Cooperation Organization, and BRICS countries; and promote the implementation of the "Mazu" intelligent meteorological early-warning solution in 30 countries, helping safeguard the lights of countless homes and protect peace and stability across the seas.

Ladies and gentlemen, friends,

As an old Chinese saying goes, "The wise adapt to changing times, and the knowledgeable respond according to circumstances." The faster AI technology develops, the more firmly we must keep it anchored to the right direction—toward the good and for the benefit of humanity; the more precisely we must calibrate the scope of regulation and governance; and the more promptly we must improve measures to prevent loss of control. We must always guide AI development with human wisdom and international consensus, so that it may truly become a powerful force for good that enhances the well-being of all humanity and advances human civilization.

China is willing, with a more open posture, more practical actions, and a longer-term vision, to work with all parties to grasp and respond to the opportunities and challenges brought by AI development, and join hands in creating an even better future for human society.

Thank you all!

Apple’s Efforts and the Regulatory Background

Since the launch of the iPhone 16 in September 2024, Apple had consistently indicated that Apple Intelligence would be available in mainland China “subject to regulatory approval.” It ultimately took nearly 22 months for Apple to complete China’s regulatory filing process. During this period, Apple navigated multiple regulatory procedures established under China’s generative AI governance framework, including model filing requirements, submission through the local Cyberspace Administration, public disclosure of approved filings, and subsequent security assessment procedures.

China’s Interim Measures for the Administration of Generative Artificial Intelligence Services were jointly issued by seven government agencies—the Cyberspace Administration of China (CAC), the National Development and Reform Commission (NDRC), the Ministry of Education (MOE), the Ministry of Science and Technology (MOST), the Ministry of Industry and Information Technology (MIIT), the Ministry of Public Security (MPS), and the National Radio and Television Administration (NRTA)—and entered into force on 15 August 2023.

The Measures establish a mandatory filing regime for providers of generative AI services that possess “public opinion attributes or the capacity for social mobilization.” In practice, service providers submit filing applications through the CAC office in their local jurisdiction. Local CAC offices publish approved filings in a timely manner, while the national CAC periodically consolidates and updates the nationwide list of approved services on its official website.

With respect to foreign companies, industry practitioners generally understand the current framework to require overseas AI providers serving Chinese users to conduct filings through an onshore legal entity under the same regulatory standards applicable to domestic providers. In addition, applicants are generally expected to provide documentation relating to cross-border data security assessments as well as compliance documentation concerning their overseas parent companies. However, the precise implementation requirements continue to depend on the specific characteristics of each project, and some aspects of the regulatory framework remain subject to case-by-case interpretation.

The major milestones between the launch of the iPhone 16 and Apple’s successful regulatory filing can be reconstructed chronologically as follows.

Regulatory Approval Is Only the First Step

From a regulatory perspective, obtaining the filing approval should be viewed primarily as an entry ticket rather than authorization for immediate deployment. Several implementation steps remain before Apple Intelligence can be officially activated on mainland Chinese devices. According to media reports, Apple must still complete the required security assessments, feature localization and adaptation, and potentially additional operating system updates before the service becomes available to Chinese users.[18]

According to Alibaba, Qwen will be integrated directly into Apple Intelligence as the underlying AI capability for users of iOS, iPadOS, macOS, and visionOS in mainland China, enabling a consistent AI experience across Apple’s entire ecosystem—including smartphones, tablets, desktop computing, and spatial computing devices. Alibaba further stated that users will be able to access Qwen’s capabilities—including text understanding, image understanding, and content generation—directly within Apple’s operating system without switching between applications. In other words, Qwen is expected to function as part of Apple’s system-level AI layer, rather than as a standalone application.

Why Apple Chose Alibaba

Apple’s decision to prioritize Alibaba over DeepSeek, Baidu, or other Chinese AI providers likely reflects multiple strategic considerations.

From a data perspective, Alibaba possesses significantly richer consumer datasets than most competitors, particularly in areas such as e-commerce behavior, digital payments, and consumer spending patterns. Training on these datasets is widely believed to improve a model’s ability to provide highly localized and personalized AI services for Chinese users.

From an engineering and infrastructure perspective, Alibaba Cloud’s localized cloud infrastructure in mainland China aligns well with the architectural philosophy of Apple’s Private Cloud Compute (PCC). Industry sources have also suggested that Qwen 3 has undergone engineering optimization for Apple’s MLX framework. Given Apple’s exceptionally high standards for model quality, reliability, and privacy protection, Alibaba may have been one of the few Chinese partners capable of satisfying Apple’s technical and operational requirements.

Alibaba as the Core LLM Provider, Baidu as the Vision Partner

Baidu appears to occupy a more specialized role within Apple’s China AI ecosystem. Rather than serving as the primary foundation model provider, Baidu is expected to supply AI capabilities for computer vision-related functions, including camera recognition and visual search.

Alibaba’s Qwen, by contrast, is expected to provide the core large language model capabilities, including natural language understanding, reasoning, and content filtering.

Why Apple Adopted a Multi-Partner Strategy

Apple’s decision to adopt a dual-supplier model, rather than relying on a single Chinese AI provider, likely reflects several strategic objectives.

First, it enables what might be described as functional specialization. Apple can separate AI capabilities into distinct layers—including perception, cognition, and compliance—and assign each layer to different domestic partners. This approach helps satisfy Chinese regulatory requirements while preserving supply chain flexibility and redundancy.

Second, Apple’s collaboration with Baidu reportedly encountered difficulties during 2024–2025, largely because Baidu’s ERNIE Bot was initially considered unable to meet Apple’s standards for model performance and privacy protection. Apple subsequently shifted toward a structure in which Alibaba became the primary LLM partner, while Baidu continued to provide specialized vision and search capabilities.

More importantly, embedding a multi-vendor switching mechanism into the system firmware allows Apple to dynamically select service providers by geography and by function. Such an architecture not only increases operational flexibility but also reduces commercial dependence on any single AI supplier.

Implications for Apple

The successful completion of the filing process is clearly a positive development for Apple.

CEO Tim Cook previously attributed part of Apple’s declining sales performance in China to the absence of Apple Intelligence. Regulatory approval now removes one of the major policy obstacles preventing Apple from closing the AI feature gap between Chinese and overseas iPhone models. This should reinforce Apple’s already strong shipment recovery in China, where iPhone shipments reportedly increased 24.4% year-over-year in the second quarter of 2026.

Nevertheless, uncertainties remain.

In June 2026, the U.S. Department of Defense added both Alibaba and Baidu to its Section 1260H Chinese Military Companies List, potentially introducing new variables into Apple’s global supply chain governance. Meanwhile, cloud-based AI functions remain constrained by China’s data localization requirements and the need to rely on domestic computing infrastructure, leaving the ultimate implementation pathway uncertain.

Moreover, regulatory filing does not automatically mean immediate commercial availability. Apple Intelligence must still complete security reviews, engineering adaptation, and possible operating system updates before deployment, and different features may ultimately be rolled out in phases.

Implications for Alibaba and Baidu

For Alibaba, the partnership represents a significant endorsement of its foundation model capabilities by the world’s most valuable consumer hardware ecosystem. It positions Alibaba Cloud as one of the default infrastructure providers through which multinational technology companies may access China’s AI market, while enabling Qwen to become deeply integrated into the system layer across iOS, iPadOS, macOS, and visionOS.

However, U.S. sanctions remain an important source of uncertainty. In addition, much of the expected commercial value from the partnership has arguably already been reflected in capital markets. Future monetization will depend largely on user adoption of Qwen-powered Apple Intelligence on Chinese devices and the resulting growth in cloud inference demand.

Baidu also benefits, although to a lesser extent. Its visual search capabilities have reportedly been incorporated into the iOS 27 Beta 2 firmware, giving Baidu an entry point into Apple’s global hardware ecosystem. This also strengthens the commercialization narrative surrounding ERNIE and Baidu AI Search by demonstrating adoption within a premium consumer hardware platform.

A Potential Template for Foreign AI Companies Entering China

Based on the currently available information, Apple’s implementation strategy suggests that on-device AI capabilities are likely to be deployed first, as they rely primarily on local computation. By contrast, cloud-based AI functions face greater uncertainty because they must comply with China’s data localization rules and operate on domestic computing infrastructure.

It is also worth noting that Samsung’s Galaxy AI was included alongside Apple Intelligence in the same batch of seven smartphone-based generative AI services approved by the CAC on 15 July 2026.

Unlike Apple’s filing, which centers on Apple Intelligence as an integrated AI platform, Samsung’s filing covers a portfolio of on-device generative AI services for mainland Chinese users rather than a single foundation model. The filing was submitted in Beijing by Samsung (China) Investment Co., Ltd.

From a technical perspective, Galaxy AI in China also adopts a multi-partner architecture. Baidu’s ERNIE provides core natural language generation capabilities, including features such as real-time call translation and text summarization; Meitu’s MiracleVision powers AI-enabled camera and image generation functions; and ByteDance’s Doubao model supports selected AI portrait and image-generation features. These third-party capabilities are complemented by Samsung’s own on-device technologies, including Samsung Knox security and local on-device inference.

More broadly, Apple and Samsung’s approach may point to a reusable compliance model for foreign AI companies or services seeking to operate in China: a domestic Chinese legal entity serves as the regulatory applicant; localised cloud infrastructure hosts cloud inference; and domestic technology partners assume a substantial share of regulatory and compliance responsibilities. Such a framework could become an increasingly important reference model for multinational AI providers entering the Chinese market.

After a lockup expiry sent the stock down more than 19%, Zhipu’s Tang Jie published a full-staff letter announcing a full return to foundation-model research and a two-year “Touch High” plan. The message is very clear: strategic investment, no chasing of short-term application revenue, resources concentrated on the underlying capabilities needed for AGI.

The framing is deliberately philosophical. Tang defines AGI not as one genius’s intelligence but as the sum of all humanity’s wisdom, and insists on continuing what the company calls its “counter-intuitive” path — the same instinct that put GLM-130B into the open before ChatGPT existed. The GLM-5.2 release is the product expression of that stance: a top-3 model on the Artificial Analysis leaderboard, shipped with a million-token context and MIT weights anyone can download, deploy, and commercialize.

Tang Jie identifies four technical peaks that must be crossed on the road to AGI.

On AI safety, Tang Jie believes capability and containment must scale together. He rejects the industry norm of treating safety as a post-hoc compliance layer, insisting instead that human ethics, social norms, and national law be embedded as foundational axioms in the model’s value function from the outset. Zhipu has committed resources in the tens of billions to “mechanical interpretability” — research aimed at converting opaque model decisions into transparent, auditable logic — and Tang cites the fact that leading overseas frontier labs have withheld full public release of their most capable models due to risk concerns as a signal, not a curiosity. His conclusion is that when a technology reaches the level of force capable of altering the course of civilization, safety is no longer an ancillary feature; it becomes the prerequisite for the technology’s permitted existence.

On Open Source, Tang frames it not as a commercial strategy but as a structural commitment to inclusive intelligence. His position is that genuine AI safety cannot be built on technological closure and barriers — it requires broad co-construction, co-sharing, and oversight conducted in the open. That conviction produced the GLM-5.2 release under the most permissive MIT license, with no restrictions based on entity type, supporting a one-million-token context and available to any developer for download, deployment, and commercialisation. For Tang, “Touch High” — pushing the frontier upward — and open access are not in tension: one hand reaches for the summit, the other paves the road down so that the heights reached belong to all of humanity, not to a handful of gatekeepers.

The Wave Has Arrived
— To every Zhipu employee and every partner who cares about the future of artificial intelligence

Allow me to use this article to address three things: who we are, how we see this era, and the strategic directions in which we have decided to invest all our strength.

(I) Who We Are: “Essence, Counter-Intuition, Focus”

Zhipu has never been a company that chases trends. It grew from a laboratory, carrying twenty years of methodology from that laboratory. This methodology can be summarized in three words: essence, counter-intuition, focus — think deeply enough, and you dare to bet against the grain; choose contrarily enough, and you must hold your ground long enough.

Looking back, almost every key decision we made once appeared “counter-intuitive.” In 2006, we sat in obscurity with an academic search system running on a single desktop computer, because we had reasoned through that behind it lay the question of “excavating the mechanisms of disciplinary evolution” — a matter worth answering over ten years. Between 2021 and 2022, when “making machines think like humans” was regarded by most as a moonshot-level fantasy, we reallocated resources, bet on hundred-billion-parameter scale, and produced GLM-130B — a full year and a half before ChatGPT set the world alight. And on the day Zhipu listed on the Hong Kong Stock Exchange on 8 January 2026, we treated it as a brand-new starting point, resolutely returning wholesale to foundational model research and driving full force toward the next generation of models.

Others rang the bell; we reset to zero. This is not a posture — it is a conviction. If the destination is AGI, then short-term interests or industry trends are merely scenery along the road to the endgame.

What has sustained us throughout is an extreme focus and a pure, unadulterated idealism. The academic search system grew from a single desktop to tens of millions of users — that took ten years. The large-model path has already consumed nearly ten years, and we will continue to cultivate it with resolve. Today’s Zhipu is a group of people willing to pursue first principles, bold enough to act counter-intuitively, and capable of seeing things through with sustained focus — that is the source of Zhipu’s core competitiveness.

(II) How We See This Era: The Upper Bound of Intelligence Is Being Rewritten

If there is one thing the past twenty years have taught us, it is this: genuine commercial opportunity never resides in minor tweaks to products or business models; it resides in the leap of intelligence’s upper bound. This is our most fundamental judgment about the current AI transformation, and the insight we most wish to convey.

This transformation is, in essence, not a product innovation or a business-model innovation — it is a technological revolution that has raised the “upper bound of intelligence.” Whoever can push that upper bound upward by even an inch first will redefine the capability frontier across every industry. All next-generation AI companies focused on first principles are competing for that inch of breakthrough.

The evolution of the intelligence upper bound follows a clear trajectory. Artificial intelligence is completing the transition from perceptual intelligence to cognitive intelligence — machines no longer merely “see” and “hear”; they are beginning to “understand” and “reason.” The next step points directly toward AGI.

We hold a plain yet exacting definition of AGI: AGI is not the wisdom of any single genius, but the sum of all human wisdom. It should be capable of creating original knowledge on the order of the Theory of Relativity — that is the only standard by which we measure whether the summit has truly been reached. On the road to that destination stand several peaks that must be crossed; they are also where today’s technological waves are most turbulent:

Peak One: Long Horizon Task
The most exciting breakthrough today is teaching models to complete extremely long tasks — not instant Q&A, but planning and execution spanning weeks, months, or even years. For example, a model can work tirelessly at the endpoint of a physics laboratory, essentially learning the thinking patterns of a top security expert and then amplifying them through the machine’s endurance.

Peak Two: Autonomous Agent System
Above long-horizon tasks, clusters of intelligent agents capable of autonomous operation, collaborative work, and 24/7 execution will become a new form of productive force. We once spoke of the “One-Person Company (OPC),” but technology is moving faster than expected — we are heading toward the “Fully Automated Company (NPC).” Memory, Continual Learning, and Self-Judging — three challenges once thought to require paradigm shifts — are gradually dissolving under the dual pressures of technology and application: long-context Retrieval-Augmented Generation (RAG) approximates the rudiment of memory; increasing model iteration frequency itself approximates continual learning; and cross-model iteration already shows early signs of self-judgment.

Peak Three: Self-Evolving
This is the most arduous and most enticing peak. AI training AI has taken shape — models write code themselves, clean and synthesize data themselves, train themselves. This may consume some compute, yet it saves the most precious resource: human labor and time. In the large-model era, speed is paramount; rapid iteration directly widens cognitive capability gaps. When leading overseas enterprises begin assembling compute clusters of one million or even two million chips, the true purpose may well be to let models train themselves.

What happens after crossing these three peaks? AI will begin to learn what “I” is — what self-cognition means; further on, it will touch human emotion; further still, consciousness itself. From perception to cognition, from cognition to generality, from generality toward Artificial Superintelligence (ASI) — this road is already laid open. The wave has come, and it is irreversible.

This is not merely our own view. Google DeepMind’s report From AGI to ASI offers a stark conclusion: even if the capability of a single model never exceeds human level, so long as compute keeps growing, superintelligence may be effectively “squeezed out.” They project that if globally operable AGI instances grow at tenfold per year, there will be one hundred million within five years. These agents — sharing the same underlying brain, thinking one hundred times more efficiently, and replicating experience at zero cost — constitute ASI at the collective level. In other words, the step from AGI to ASI requires both algorithmic breakthroughs and the aggregation of massive compute resources.

This irreversible trend will penetrate the entire technology stack from top to bottom: when AGI arrives, today’s applications may all need to be reconstructed as AI-native, or may no longer be needed at all; operating systems may be rewritten — in the future, when you open a computer, you will see an “LLM OS” with all functions generated on demand; deeper still lies a challenge to the von Neumann architecture that has run for eighty years. Finance, law, e-commerce, the internet — no industry will be exempt. Many friends have come to me saying they want to transform their enterprises and keep pace with AI, yet those who have truly grasped that “this irreversible transformation has already begun” remain few.

(III) The Direction in Which We Invest All Our Strength: “Touch High”

Once the trend is clear, what remains is choice. And Zhipu’s choice is, as always, “counter-intuitive” — at a moment when the industry is broadly accelerating commercial monetization, we have decided to break upward.

We have named this strategy the “Touch High Plan.” At the historic inflection point where artificial intelligence transitions from perception and cognition toward fully general intelligence, Zhipu will adopt a “touch high” posture to challenge the physical and algorithmic limits of current technology. Over the next two years, we plan to invest strategically — not pursuing short-term application monetization, but aiming directly at AGI’s next high ground.

This investment will be concentrated on four core engines:

Engine One: Long Horizon Task. Moving AI from “instant Q&A” to “grand-scale engineering” — developing next-generation memory architectures that enable models to “learn, act, and remember” continuously throughout a project’s entire lifecycle, and to autonomously decompose grand objectives (such as “design a novel anticancer drug molecule”) into thousands of executable sub-tasks.

Engine Two: Autonomous Agent System. Moving from “intelligent assistant” to “digital employee” — constructing an agent society containing tens of thousands of agents with different professional “personalities” and “skills,” enabling them to debate autonomously, collaborate, review code, and schedule resources, achieving “autonomous-driving”-level digital productivity.

Engine Three: Fully Self Training. As high-quality human data approaches exhaustion, converting compute into fuel for evolution — building high-quality synthetic data factories, achieving the “creation of knowledge from nothing” through AI-versus-AI adversarial Self-Play, and granting systems the capability to restructure their own code within secure sandboxes, so that the pace of evolution breaks free from the physical constraints of human engineers.

Engine Four: Extreme Safety Governance. This is the one among the four engines I most wish to emphasize.

The stronger the capability, the more robust the safety-constraint mechanism must be. From its founding, Zhipu established a guiding principle: AI must serve human well-being and national strategy. The company rejects bolt-on safety patches and insists on embedding human ethics, social norms, and national laws and regulations as foundational axioms in the model’s value function. We plan to commit resources in the tens of billions to tackle “mechanical interpretability,” clarifying the neural logic behind model decisions and advancing the transformation of black-box systems into transparent, explainable systems; we also actively participate in international AI governance to prevent the misuse of AI technology.

This urgency is not unfounded anxiety. When the most advanced overseas frontier models have deferred full public release due to risk considerations, and their corporate leaders have publicly warned that AI’s far-reaching impact will profoundly reshape the global balance of power, we should be all the more clear-eyed: the realization of superintelligence and the research into superintelligent containment must advance in parallel. This is also a proposition we examine repeatedly when facing forward-looking technology — history has repeatedly demonstrated that when a technology reaches the level of force capable of altering the course of civilization, safety is no longer an ancillary feature; it becomes the fundamental prerequisite for the technology’s continued existence and permitted application.

(IV) Open Ecosystem: The Underlying Logic of Inclusive Intelligence and Safety Governance

We have always believed that artificial intelligence, as a strategic technology leading the future, cannot develop over the long term without an open and collaborative industrial ecosystem. The value of frontier intelligence lies not only in the technological breakthrough itself, but in whether it can broadly empower all industries and benefit every developer. We are convinced that genuine safety is not built on technological closure and barriers, but on broad co-construction, co-sharing, and oversight conducted in the open.

It is precisely from this deep recognition of technological inclusivity that Zhipu has formulated its strategic response. Recently, we released GLM-5.2 — our strongest open-source model to date — supporting a truly usable one-million-token (1M) context, maintaining leadership in long-horizon tasks, open to all users, and to be formally open-sourced under the most permissive MIT license. Anyone can download, deploy, and commercialize it, with no restrictions based on entity type. This is the company’s resolute stance expressed in product form.

We have chosen to believe in another path: frontier intelligence should not belong only to a few, nor should it be revocable by a few rules at any time. It should be open, usable, buildable, and serve every developer.

This does not contradict “Touch High” — it is the other side of the same coin: one hand reaching up to touch the heights, challenging the limits of intelligence; the other hand paving the road downward, making the most cutting-edge capabilities as open and inclusive as possible. The heights reached belong to all of humanity; the road built belongs to every individual.

(V) Conclusion: Why Now, Why Us

Some will ask: why, after listing, does Zhipu continue to pour its core resources into the most uncertain direction of “touching high”? Because we believe a simple truth: those who truly reach the summit turn the mountain into a road.

The essence we clearly perceived was once crystallized, through the “WuDao Large Model” project, into the shared conviction of hundreds of scientists; and then, through Zhipu’s industrial investments and its entire ecosystem, it became the foundation from which a generation of entrepreneurs could take their leap. Today, we want to build this road higher and wider — high enough to protect ourselves and safeguard national security; high enough to give humanity the chance to explore more unknowns and the mysteries of the universe; and wide enough for every developer and every team to walk upon it.

In the AGI era, these once-unreachable aspirations have, for the first time, a genuine prospect of realization. This is the greatest fortune of our generation of Chinese people — and also the heaviest responsibility.

The wave has come; the trend is irreversible. Zhipu will be the one who meets the crest of the wave and reaches upward.

Not reaching the summit is failure.

This time, what we intend to touch is a height that belongs to all of humanity.

Zhipu Founder, Tang Jie
11 July 202

Among China’s large-model founders, Zhipu’s leaders have been unusually explicit — and unusually consistent — in tying the company to AGI since 2019. The vocabulary, the roadmap, and the refusal to pivot all predate the July 2026 letter by years. Read together, the public statements explain why “Touch High” was less a reversal than a return.

“We don’t build China’s ChatGPT”

Zhang Peng set the position early and has not moved off it. In a 2023 interview he explained why Zhipu refused to sell vertical, industry-specific models: “an industry model is essentially rebuilding the wheel of traditional algorithms inside the shell of a large model,” and “only a general model of a certain scale can produce human-like cognitive emergence”. The framing is a business decision dressed as a technical one — betting the company on a general base model rather than the easier services revenue.

He has repeatedly named AGI as the terminal objective, not a marketing flourish. Zhang has said that if Zhipu abandoned base-model training to merely call others’ APIs, or traded the AGI goal for near-term monetization, the company would “lose its meaning” to him. Asked directly whether AGI is achievable, he told an English-language interviewer that AGI “has never had a fully authoritative definition,” but “as long as we define the goal properly — not in some absurd way — then in all likelihood it can be achieved. The real question is how long it will take”; AGI, he added, has been the goal “from the first day Zhipu was founded”.

The five-level map to AGI

Zhipu is the Chinese firm that has most openly published an AGI roadmap. Tang Jie’s five-level scheme, which Zhang Peng also uses publicly, defines the climb.

Zhipu places today’s frontier around L1 to L3, which is why agents, reasoning, and self-improving training keep recurring in its releases. The slogan on the wall states the same ambition in five characters: “让机器像人一样思考” — make machines think like humans. Chairman Liu Debing invoked it at the Hong Kong listing, noting that “Z” is the last letter of the alphabet, the ultimate state, and reading the ticker 2513 as “AI, my whole life”.

Reframing AGI as a business

The one place Zhang Peng translates the creed into unit economics is the capital-markets story. After Zhipu’s first post-IPO results in March 2026, he offered what he called the “first principle” of AGI commercialization: AGI commercial value equals the intelligence ceiling multiplied by token consumption. The intelligence ceiling determines pricing power; token scale determines the size of the value pool; and the keyword for 2026, in his telling, is “token volume”. The formula is how a research-first founder justifies a technology multiple to shareholders without conceding the research priority.

Tang Jie’s radicalism

Tang Jie supplies the harder edge. He halted the team’s internationally recognized graph-neural-network and knowledge-graph research — work with real academic standing — and moved the whole group to large language models before the bet was safe After DeepSeek’s emergence in early 2025, he judged the “chat paradigm” to be near its ceiling and reframed the goal as “making AI actually do things,” folding coding, agentic, and reasoning capabilities into GLM-4.5. His theoretical stance is deliberately provocative: “domain-specific large models are a false proposition,” and online learning plus self-evaluation are the new scaling paradigm rather than ever-larger static pre-training. His summit rule is the blunt version of the same idea: “reach the summit, or it is failure”.

Strip away the stock drama and Zhipu is running a live experiment on three questions that matter well beyond one company.

• Can open weights be a durable strategy at the frontier? Zhipu is betting that giving away SOTA models builds a moat through developers and ecosystem rather than eroding one. The 700,000-developer base is evidence for; the $105 million revenue against a $128 billion valuation is the evidence against.

• Does an Entity List placement bind a software company? So far the designation has reshaped Zhipu’s supply chain toward domestic chips without visibly slowing its model releases — a different outcome than hardware-centric designations.

• Can a research-first culture survive public markets? The Touch High plan is a direct refusal of the quarterly logic that public listing usually imposes. Whether shareholders tolerate a two-year monetization pause after a 2,200% run is the open question the next eight quarters will answer.

The company’s own history is the reason to take the gamble seriously. The last time Zhipu made a large, unpopular, long-horizon bet — trillion-scale open pre-training in 2021–2022 — it was early rather than wrong. Touch High asks the market to extend the same benefit of the doubt one more time.

That makes this July 10 announcement much more interesting.

Announcement No. 29 of 2026 of the Ministry of Commerce and the General Administration of Customs on the Implementation of Temporary Export Prohibition Controls on HeliumIssuing Authorities: Ministry of Commerce of the People's Republic of China; General Administration of Customs of the People's Republic of China

Document No.: Announcement No. 29 of 2026

Issuance Date: July 10, 2026

Commodity: Helium | Customs Commodity Code: 2804290010

Effective Date: Effective from the date of promulgation

Pursuant to the relevant provisions of the Foreign Trade Law of the People's Republic of China, the Ministry of Commerce and the General Administration of Customs have decided to implement temporary export prohibition controls on helium (Customs Commodity Code: 2804290010).This Announcement shall take effect as of the date of promulgation. Any subsequent adjustments will be announced separately.Ministry of Commerce / General Administration of Customs

July 10, 2026

The MOFCOM announcement itself is extremely short. It just said helium under customs code 2804290010 would be subject to a “temporary export prohibition” from July 10. There is no stated expiry date, no licensing process, no country exemptions, purity thresholds or transition arrangements for existing contracts.

This is also legally different from China’s better-known export controls on dual-use items. The notice relies only on the Foreign Trade Law. It does not cite the Export Control Law, the Customs Law or the regulations governing dual-use exports. This is looks less like a national-security licensing regime and more like emergency management of a product in short supply.

Helium is easy to dismiss as the gas used in party balloons. In reality, its most important applications are highly strategic. Helium is chemically inert, extremely light and highly conductive. Liquid helium also remains liquid at temperatures close to absolute zero. These properties make it difficult to replace in MRI machines, rocket systems, advanced scientific research and semiconductor manufacturing.

And Helium matters to chips. In a chip fab, helium performs several jobs. Its most important role is wafer cooling. During plasma etching and other high-energy processes, heat builds up on the silicon wafer. A thin layer of helium is introduced between the wafer and the electrostatic chuck underneath it, allowing heat to move away quickly and evenly.

If temperature control becomes unstable, the wafer can deform, process uniformity can deteriorate and yields can fall.

Helium is also used as a carrier or purge gas, in ion implantation, in some deposition and lithography equipment, and for highly sensitive leak detection in vacuum systems.

Alternatives exist for some of these uses. Nitrogen or argon can replace helium in certain processes. New cooling technologies are also under development. But switching gases inside an advanced production line is not like changing suppliers for office equipment. It can require tool modifications, process qualification and months of testing.

For the most demanding steps in advanced chip manufacturing, there is no substitute that fabs can immediately deploy at scale.

That does not mean a helium shortage shuts down a fab overnight. Large chipmakers usually hold inventories, sign long-term contracts and receive priority from industrial-gas suppliers. During a shortage, party balloons, welding and other lower-value uses are cut first.

But once inventories fall below the minimum operating needs of critical equipment, fabs may have to slow production, prioritize their most valuable products or temporarily idle individual tools.

For a country spending heavily to expand semiconductor capacity, that is not a trivial risk.

The helium market is small, opaque and extraordinarily concentrated.

According to the [US Geological Survey](https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-helium.pdf), global helium production reached roughly 190 million cubic metres in 2025. The United States produced about 81 million cubic metres and Qatar another 63 million. Together, they accounted for roughly three-quarters of global output.

Russia produced around 18 million cubic metres and Algeria 11 million. The four largest producers therefore controlled more than 90% of the market.

This concentration is only part of the problem. Most helium is recovered as a by-product of natural-gas processing. A country can possess substantial underground resources without being able to bring commercial helium to market. Production depends on the economics of the gas field, the concentration of helium in the raw gas, purification and liquefaction facilities, and access to specialized cryogenic containers.

The containers themselves are a bottleneck. Liquid helium must be kept at extremely low temperatures and gradually evaporates during transport. Containers must then be returned to the producer for refilling. When shipping routes are disrupted, the shortage of containers can become as important as the shortage of helium.

This is exactly what happened in 2026. The war with Iran disrupted Qatar’s LNG operations and shipping through the Strait of Hormuz. Because Qatar produces helium alongside LNG, the disruption temporarily removed close to one-third of global supply from the market. Damage to Qatar’s facilities is expected to reduce its helium export capacity by around 14% even after partial operations resume.

Russia could have provided some relief. Instead, Moscow introduced its own restrictions in April. Helium exports outside the Eurasian Economic Union now require special government approval, with the regime scheduled to remain in place until the end of 2027.

The European Union, meanwhile, has prohibited imports of Russian helium since 2024 as part of its sanctions regime.

This led to a global shortage as well as a regional fragmentation of the helium market. North American production increasingly supports the United States and its partners. Russian helium is pulled towards China and Eurasia. Qatar has less flexible supply available beyond its most important long-term customers. Industrial-gas companies are allocating scarce volumes to medical, aerospace and semiconductor users.

Helium has become another market in which physical scarcity and geopolitical alignment reinforce each other.

China has some helium resources, mostly associated with natural-gas deposits. But these resources are often low-grade, scattered and expensive to extract.

Domestic production has grown rapidly from a very low base. China had nominal production capacity of roughly 14.7 million cubic metres in 2025, but actual output was only around 4.6 million cubic metres. Many facilities operated far below designed capacity.

China therefore remained heavily dependent on imports. In 2025, it imported more than 4,900 tonnes of helium while producing less than 900 tonnes domestically. Depending on the calculation method, its import dependence was around 84% to 87%.

The sources were also highly concentrated. Qatar and Russia together supplied approximately 98% of China’s helium imports. This was an improvement over China’s earlier reliance on Qatar alone. Russia’s Amur gas-processing complex gave China a second major source that could be transported overland rather than through the Strait of Hormuz.

But diversification from one supplier to two is not genuine supply-chain resilience—especially when one is exposed to a Middle Eastern war and the other has just introduced export approvals.

China’s vulnerability is therefore larger than the import-dependence number suggests. It relies not only on foreign resources, but also on overseas liquefaction facilities, industrial-gas companies, specialized containers and cross-border transport.

China had also begun to function as an Asian processing and trading hub. During the first eleven months of 2025, China imported around 4,294 tonnes of helium and exported approximately 438 tonnes. Some of this was helium purified or processed in China. Some likely reflected regional trading and the re-export of Russian-origin supply.

The exported volume was not huge. Converted into gaseous helium, it amounted to roughly 2.6 million cubic metres—only around 1% to 2% of annual global output.

But in a tight and illiquid market, marginal volumes matter. The impact of removing 2.6 million cubic metres from the Asian spot market can be much larger than the headline global percentage suggests.

China consumed roughly 4.4 million cubic metres of helium in semiconductor manufacturing in 2024, about 16% of its total demand. That share is likely to rise as China builds more wafer fabs, expands advanced memory production and pushes deeper into AI chips, commercial space and high-end scientific equipment.

Helium will not determine whether China catches up with the United States in advanced semiconductors. Lithography equipment, high-bandwidth memory, electronic-design software and leading-edge manufacturing knowledge remain much larger constraints.

But modern chip production is a system. A fab cannot operate with 99 of the 100 inputs it needs. The most advanced lithography machine is useless if the factory cannot secure the gases required to run the surrounding processes reliably.

The ban therefore does three things at once.

First, it keeps Chinese-produced helium inside the country.

Second, it prevents imported helium from being re-exported when international prices rise above domestic prices.

Third, it gives China greater control over how scarce domestic supply is allocated among chipmakers, hospitals, aerospace programmes, research institutes and lower-priority industrial users.

But generally, this is domestic supply management, not resource coercion that many western nations would assume.

The immediate international impact of the ban will fall mainly on Asian users that obtained Chinese-processed helium or Russian helium routed through China.

South Korean, Japanese and Taiwanese semiconductor companies may need to seek additional supply from the United States, Canada or Algeria, or obtain Russian helium through more direct and politically complicated channels.

Russia will lose some flexibility to reach third-country buyers through China, but its importance to the Chinese domestic market will increase. That may strengthen Beijing’s bargaining position in long-term contracts and overland logistics.

For global industrial-gas companies, the ban removes another source of flexible supply. Long-term customers will continue to receive priority, while smaller buyers face higher prices, surcharges and deeper allocation cuts.

The lowest-priority uses will suffer first. Semiconductor fabs and MRI systems will probably remain near the front of the queue.

Li reaffirmed China’s support for the United Nations as the primary platform for global AI governance. He called for continued efforts to bridge the global AI divide by helping developing countries, particularly those in the Global South, strengthen their AI capabilities. He also advocated for greater open-source innovation to advance technological progress and broaden the benefits of AI, and urged the international community to jointly build a global AI governance framework based on broad consensus, including common standards and governance principles.

At the same time, the Chinese government also submitted a written submission to the Dialogue.

Notably, Li also made China’s position unmistakably clear: the Chinese government strongly supports open-source AI, explicitly citing DeepSeek and Qwen as leading examples.

Open source AI is a shared asset for all humanity. Chinese open source models such as DeepSeek and Qwen have significantly lowered the barriers and costs of AI adoption. China is committed to further promoting open source AI for industry, academia and research institutions, encouraging innovation, AI empowerment and an inclusive ecosystem through international cooperation, thereby injecting sustained momentum into AI development.

Interestingly, there have recently been report that China is considering export controls on open-source AI models. A little-noticed roundtable hosted by the Supreme People’s Court—one that virtually no one in China’s policy community paid attention to—has also been brought up again in US.

Now that you’ve seen the Chinese government’s position on open-source AI, do you still think Beijing is going to impose export controls on open-source models?

And frankly, that roundtable discussion was largely irrelevant.

Speech of Li Lecheng, Minister of China’s MIIT:

Your Excellencies, President Ms. Baerbock, UN SG Mr. Guterres, co-chairs, Heads of State and international organizations, colleagues, AI governance concerns the future of humanity and is a shared task of all countries. Chinese President Xi Jinping proposed Global AI Governance Initiative, calling for inclusive, equitable and effective AI governance to ensure AI for good. China has consistently supported the UN as the primary channel for global AI governance and has worked to help bridge the global AI divides. To this end, I’d like to make three proposals.

First, AI for all. Regardless of their size, strength or social system, all countries have the equal right to develop and benefit from AI. China has launched the AI capacity building action plan for good and for all and implemented concrete projects to promote South-South and triangular cooperation. China stands ready to make full use of the group of friends for international cooperation on AI capacity building to actively support other countries, especially those in the global South, in developing AI capabilities and sharing the AI dividend.

Second, open source cooperation and innovation. Open source AI is a shared asset for all humanity. Chinese open source models such as DeepSeek and Qwen have significantly lowered the barriers and costs of AI adoption. China is committed to further promoting open source AI for industry, academia and research institutions, encouraging innovation, AI empowerment and an inclusive ecosystem through international cooperation, thereby injecting sustained momentum into AI development.

Third, jointly building global AI governance. AI presents tremendous opportunities but it also brings unforeseen risks and challenges. China has therefore put forward the Global AI Governance Action Plan, firmly upholding the principle of balancing between development and safety, advocating technical ethics review and other governance proposals to ensure AI for good. China stands ready to work closely with all parties to advance global AI governance frameworks, standards and norms based on broad international consensus.

Dear colleagues, the development and governance of AI will shape our common future. China calls on all parties to strengthen solidarity to promote AI for good and for all.

From 17th to 20th July, China will host the World AI Conference and High-Level Meetings on Global AI Governance in Shanghai. I sincerely invite you to join us in person and contribute to advancing global AI development.

Thank you.

Chinese government’s official submission to the UN Geneva AI conference:

Submitter: Permanent Mission of China to the UN

Respondent ID: 1605

Region: Global

Stakeholder Category: Government

Q1. The Global Dialogue on Artificial Intelligence should be guided by the principles of "inclusive and shared benefits, open and collaborative efforts, safe and controllable development, and contribution to sustainable development." Special emphasis should be placed on strengthening AI capacity-building and addressing the common concerns of the Global South. The Global Dialogue should serve as an important platform for policy communication, experience sharing, and international cooperation.

Q2. AI capacity-building; Social, economic, ethical, cultural, linguistic and technical implications of AI; Interoperability of governance approaches; Open-source software, open data and open AI models.

Q3. The Global Dialogue on AI should be a forward-looking, action-oriented, and truly inclusive process. It must go beyond technical discussions to address power asymmetries, capacity gaps, and long-term risks. By centering the needs of the Global South, championing openness, and embedding safety and sustainability from the start, the UN can help shape an AI future that is shared, cooperative, secure, and sustainable for all.

Q4. Contributing to Sustainable Development – Long-Term and Holistic Perspective:

- AI should be developed and deployed in a green and low-carbon manner to support global climate and environmental goals.

- Member States must also take into account the long-term and systemic impacts of AI, including on energy consumption, ecological systems, employment structures, and social equity.

Q5. Safe and Controllable Development – Managing Risks:

- AI governance must balance development and security, ensuring innovation proceeds without exposing societies to unacceptable risks.

- A tiered and category-based risk management approach should be adopted, tailored to different contexts and use cases.

- Particular attention should be paid to preventing potential abuse of AI technologies, including in surveillance, disinformation, autonomous weapons, and fraud.

Q6. Inclusive and Shared Benefits – Bridging the AI Divide:

- The development of AI should aim to advance the shared well-being of humanity. It must adhere to the principles of fairness, inclusiveness, and non-discrimination.

- International efforts should help all countries — especially those in the Global South — close the intelligence divide and share the dividends of AI development.

- No country or community should be left behind or systematically excluded from AI's potential benefits.

Q7. Open and Collaborative Efforts – Opposing Fragmentation:

- All countries have the equal right to access, develop, and deploy AI according to their own development paths and national conditions.

- The Global Dialogue should oppose drawing ideological lines, forming exclusive blocs, or fragmenting global industrial and supply chains under the guise of AI governance.

- Greater interoperability of AI governance frameworks, norms, and standards should be actively pursued to reduce friction and enable cross-border cooperation.

The document follows a clear lifecycle-based security framework. It requires AI agents to undergo security assessments before they are put into use, complete security hardening prior to deployment, operate under strict permission controls throughout their lifecycle, and ensure that all data is securely erased when they are decommissioned.

Across every stage of this lifecycle, the guidelines repeatedly emphasize six practical security requirements: ensuring the integrity and trustworthiness of software and model sources; enforcing the principle of least privilege; minimizing network exposure; maintaining comprehensive audit logs of agent activities; implementing controls over high-risk operations; and applying enhanced safeguards to sensitive data and long-term memory.

The guidance suggests that China no longer regards AI agents merely as lightweight application-layer components built on top of large language models. Instead, regulators increasingly view them as integrated systems that combine memory, tool use, autonomous decision-making, and operational privileges—systems that introduce an entirely new category of security risks requiring dedicated governance.

Separately, the Cyberspace Administration of China, together with four other government agencies, has issued the Interim Measures for the Administration of Anthropomorphic AI Interaction Services, which will take effect on 15 July 2026.

In early July, both Doubao and Tongyi Qianwen notified users that certain AI agent features would be discontinued during the same period. According to multiple media reports, these adjustments form part of a broader compliance campaign aimed at bringing anthropomorphic AI services and user-created AI agents into line with the new regulatory framework. Importantly, this should not be interpreted as China seeking to restrict AI agent technology as such. Rather, it reflects a shift toward more granular regulation, with authorities requiring major platforms to remove or modify product features with ambiguous regulatory boundaries before the new rules formally come into force.

Below is the full translation of the standard:

TC260-PG-20266A

Cybersecurity Standard Practice Guide— Security Guidance for Agent Deployment and Use

(v1.0-202607)

Secretariat of the National Technical Committee 260 on Cybersecurity of Standardization Administration of China

July 2026

This document is available at:www.tc260.org.cn/

Foreword

The Cybersecurity Standard Practice Guide (hereinafter referred to as the “Practice Guide”) is a standards-related technical document organized, developed, and issued by the Secretariat of the National Technical Committee 260 on Cybersecurity of Standardization Administration of China (hereinafter referred to as the “TC260 Secretariat”). It is intended to promote cybersecurity-related standards and knowledge and provide standardized practice guidance around topics such as cybersecurity laws, regulations and policies, standards, and cybersecurity hotspots and incidents.

This document was drafted by the China Electronics Standardization Institute, National Computer Network Emergency Response Technical Team/Coordination Center of China, Tsinghua University, Pujiang National Laboratory, Beijing Zhongguancun Laboratory, Data and Technology Support Center of the Cyberspace Administration of China, Zhejiang University, Alibaba Cloud Computing Co., Ltd., Huawei Cloud Computing Technologies Co., Ltd., Beijing Volcano Engine Technology Co., Ltd., Beihang University, China Mobile Communications Group Co., Ltd., The Fifth Electronics Research Institute of the Ministry of Industry and Information Technology, Beijing University of Posts and Telecommunications, Beijing Shengxin Network Technology Co., Ltd., Tencent Cloud Computing Co., Ltd., Hangzhou Anheng Information Technology Co., Ltd., Fangcun Yueqian (Xiong'an) Technology Co., Ltd., Hangzhou Turing Zhiyue Technology Co., Ltd., Beijing NSFOCUS Technologies Group Co., Ltd., and Beijing Qihoo Technology Co., Ltd.

The principal drafters of this document are Yao Xiangzhen, Bo Zhaoyi, Ren Kui, Xu Wei, Zhou Ruikang, Wang Bo, Wang Yingchun, Cui Yong, Li Qi, Hao Chunliang, Wang Zhiwei, Liu Xianglong, Qin Zhan, Guan Zhenyu, Zhang Lei, Zhang Yanting, Xu Yang, Li Ziwei, Meng Lingyu, Zhao Yuhang, Cui Dong, Chu Zhixuan, Shi Guixin, Peng Juntao, Geng Tao, Shao Meng, Guo Jianling, Yuan Kaiguo, Li Xiaojian, Wang Jiakai, Zhao Bin, Fan Chunpeng, Zou Quanchen, Su Ruisheng, Sun Lipeng, Xiao Na, Chen Xing, Ye Xiaohu, Zheng Yanting, Yu Xionghui, Hu Hao, Chen Yanxu, Zhong Yucheng, Yang Shaobo, and Peng Xiaojun.

Statement

The copyright of this Practice Guide belongs to the TC260 Secretariat. No part of the Practice Guide may be copied or translated in any manner without the written authorization of the Secretariat.

When reproducing or citing the views or data in this Practice Guide, please indicate the source as: “Secretariat of the National Technical Committee 260 on Cybersecurity of Standardization Administration of China”.

1 Scope

This document provides security guidance for the deployment and use of agents, covering the stages of assessment, preparation, deployment, use, and decommissioning.

This document applies to the prevention of security risks in the deployment and use of agents, and may also serve as a reference for selecting and using commercial agent services.

2 Normative References

The contents of the following documents constitute indispensable provisions of this document through normative reference in this text. For dated references, only the version corresponding to that date applies to this document. For undated references, the latest version, including all amendments, applies to this document.

GB/T 25069 Information Security Technology Vocabulary

3 Terms and Definitions

The terms and definitions defined in GB/T 25069 and the following apply to this document.

3.1 agent

An intelligent system with capabilities for autonomous perception, memory, decision-making, interaction, and execution.

Note: In this document, “agent” specifically refers to an AI agent based on large models, oriented to personal-assistant scenarios, and requiring the user to grant relatively high privileges. It is generally a software system.

3.2 tools

Standardized interfaces that enable agents to invoke external capabilities, defining the name, input parameters, and return results of an atomic operation.

3.3 skills

A standardized instruction set for specific tasks, encapsulating best practices, operating procedures, and domain knowledge for completing a certain type of task.

4 Abbreviations

API: Application Programming Interface

IP: Internet Protocol

5 Overview

During the deployment and use of agents, users should carry out security protection in accordance with the security guidance in Chapters 6 to 10 of this document, and perform security checks with reference to Appendix A. If security issues are identified during the security check or in the course of use, users should promptly strengthen security according to the relevant chapters of this document. Users shall carry out agent deployment and use activities in compliance with the regulations of the state and of their organizations. To address various security risks that may arise from internal personnel deploying and using agents, relevant organizations should establish corresponding security management systems and supporting security capabilities in accordance with the security management guidance in Appendix B.

6 Assessment Stage

This stage mainly involves assessments prior to agent deployment and use, including the necessity of using agents, the applicability of open-source agents and commercial agents, and whether the proposed agent to be selected meets the baseline security feature requirements.

a) The purpose of use, business scenario, and work task should be clearly defined, and the necessity and reasonableness of using an agent to complete the relevant tasks should be assessed.

b) The user should understand in advance the technical characteristics, capability boundaries, and security risks of agents, as well as the principal differences between open-source agents and commercial agents.

c) Based on the needs analysis in a) and the knowledge preparation in b), the agent to be deployed and used should be selected prudently. Priority should be given to complete solutions in which the provider offers both the agent and supporting security protection capabilities. Open-source projects lacking security protection measures should be selected with caution.

Note: When selecting a commercial agent, the user should carefully read the security precautions issued by the provider and verify its security features against the subsequent chapters of this document. If the requirements are met, it is unnecessary to repeat the subsequent security measures. When selecting an open-source agent for which security protection must be implemented independently, security protection should be strengthened in accordance with the subsequent chapters of this document.

d) The maintenance status of the proposed agent project should be checked. An agent project shall not be selected if any of the following conditions exists:

1. There has been no maintenance record for more than one month, and unresolved security issues exist.

2. Publicly disclosed security issues have gone unanswered for a long period.

3. Unpatched high-severity vulnerabilities exist.

4. The provider has explicitly ceased maintenance.

e) Priority should be given to agents that integrate mechanisms such as secure sandboxes, high-risk operation control, emergency stop control, and traceable file systems. Agents lacking basic security mechanisms such as log auditing and permission management shall not be selected.

f) The proposed agent to be deployed and used should be checked for major security risks such as automatically opening public network interfaces or forcibly returning runtime data.

7 Preparation Stage

This stage mainly concerns preparation for the installation and deployment of the agent, including installation materials, deployment environment, large models, and security protection.

a) When selecting installation materials for the agent:

1. Agent installation files, container images, and other installation materials should be obtained from channels with clearly identified responsible entities and release mechanisms, such as the provider’s official website or application stores. Modified versions, redistributed versions, cracked versions, or other installation materials of unknown origin from network drives, forums, instant-messaging groups, and similar channels shall not be used.

2. Before installing the agent, the authenticity and integrity of installation materials should be verified through the digital signatures, hash values, checksums, or similar methods provided by the releasing entity, in order to prevent poisoned or tampered installation materials.

b) Corresponding security protection measures should be taken according to the installation and deployment method of the agent:

1. Where deployment is to be carried out directly in a local environment, a dedicated device should be chosen for installing the agent, and it should not be mixed with devices used for daily life, work, office work, and similar purposes. Before deployment, the existing files on the device should be backed up. If the device contains sensitive data or private files, such data or files should be cleaned up or migrated in advance.

2. Where deployment is to be carried out in a virtualized environment such as a virtual machine or container, strict isolation should be implemented through configurations such as prohibiting the virtual machine from accessing files on the host machine, so as to prevent security issues from spreading to the host.

3. Where deployment is to be carried out in a cloud environment, a cloud platform with capabilities such as agent identity management, access control, log auditing, and security alerting should be selected. Before deployment, the cloud service provider should be consulted to confirm whether the security capabilities it provides meet the relevant requirements of this document.

c) When selecting the large model required for agent operation:

1. A large model that has completed the filing required for generative AI services should be selected.

2. Where there are relevant requirements for data security and privacy protection, and the necessary computing conditions and technical foundations are available, a locally deployed large model should be given priority.

3. Where an external large-model service is invoked, the official interface of the large-model service provider shall be used. A large model shall not be invoked through relay services of unknown origin or unauthorized agents.

4. A large model whose context window, capability boundaries, and security level meet the requirements should be selected according to the complexity of the agent task and the security needs, so as to avoid agent misoperation caused by insufficient context window length or inadequate security capabilities.

d) According to security needs, a gap analysis should be conducted on the existing security mechanisms of the selected agent and large model. Where the following security capabilities are insufficient, they should be strengthened by deploying security tools, connecting security services, or similar means:

1. Security protection for input and output content.

2. Agent behavior control, such as runtime behavior monitoring and anomaly circuit breaking.

3. Security management of agent identities and credentials.

4. Security inspection of the agent supply chain, such as skills and tools.

5. Interception and management of high-risk operations.

6. Environment isolation based on technologies such as sandboxes.

7. Resource consumption and cost control.

8 Deployment Stage

This chapter sets out security guidance for the deployment stage of the agent, including deployment methods, plugin installation, runtime accounts, permission configuration, network exposure control, log auditing, and management of high-risk operations.

a) Deployment should be carried out according to the official documentation of the agent provider or deployment scripts that have been verified. Convenience scripts such as one-click deployment scripts from unknown sources shall not be used.

b) Where plugins need to be installed additionally, the reliability of the plugin source should be checked before installation. Plugins from unknown sources, with unpatched high-severity vulnerabilities, with abnormal permission requirements, or unrelated to the intended task shall not be installed.

c) The agent shall not be run with operating-system administrator privileges. Only the minimum permissions necessary for the agent to complete the intended task should be granted, so as to reduce the impact range of agent overreach or abuse.

d) The range of directories accessible to the agent should be restricted. The agent shall not be granted default access to important locations such as the user’s home directory or system directories. A dedicated working directory should be established for the agent, and only the files necessary for the task should be placed in that directory.

e) The agent and its related services should be configured to be accessible only from the local machine and not exposed to external networks, in order to prevent unauthorized remote invocation. Where access to the agent through the internet is genuinely required, encrypted access shall be configured and access sources shall be restricted.

f) The logging function of the agent should be enabled. Full logs should be kept for actions such as file operations, instruction execution, network activities, and skill invocation. The logging granularity should be set to the finest level supported.

g) A list of high-risk operations should be established in advance and managed through the agent’s security mechanisms. Operations on the list should be subject to secondary confirmation or direct blocking. High-risk operations include, but are not limited to, stopping or disabling system services, terminating processes, formatting or overwriting disks, bulk deletion, permission modification, key changes, opening ports, modifying firewall rules, modifying system settings, and accessing password managers. According to the purpose of agent use, operations involving transfers, payments, and other acts affecting personal property security should be included in the list of high-risk operations.

9 Use Stage

This chapter sets out security guidance for the use stage of the agent, including security re-check, skill security, sensitive information protection, network access security, long-term memory management, emergency response, and updates.

a) The installation and configuration of the agent should be reviewed, including but not limited to the following:

1. Whether the user can effectively understand the execution status of the agent.

2. Whether the user can carry out security management over the operations executed by the agent, for example by configuring blacklists and whitelists.

3. Whether the user can conveniently trigger emergency stop control of the agent.

b) Security protection for the use of skills should be strengthened, including but not limited to the following:

1. Skills from reliable sources shall be used.

2. Comprehensive security testing of skills should be carried out. Where skills that have not undergone security testing are used, risk precautions should be prepared in advance.

c) When using an agent connected to the internet, attention shall be paid to protecting the security of sensitive data, including but not limited to the following:

1. When providing personal information to the agent, the principle of minimum necessity shall be followed. Sensitive information involving biometric characteristics, family privacy, and the like should be provided with caution.

2. Without third-party authorization, the personal information and sensitive data of third parties shall not be provided to the agent.

3. Business data for which the user has not obtained permission, and content involving intellectual property, shall not be provided to the agent.

d) Where the agent is accessed through the internet, the necessity and security of its public network interfaces should be checked regularly, and non-essential exposed interfaces should be closed in a timely manner.

e) Where the agent is used to obtain online information services or participate in online social-media discussions, the relevant cybersecurity laws and regulations shall be complied with.

f) Valuable data in the environment in which the agent operates should be backed up regularly for disaster recovery.

g) The contents recorded in long-term memory files should be reviewed manually on a regular basis. Personal information, private data, and internal information that should not be stored should be handled in a timely manner, and the long-term memory configuration strategy should be adjusted accordingly.

h) Sensitive permissions should be revoked in a timely manner, unused skills and sensitive conversation records should be cleaned up regularly, and multi-factor authentication should be enabled to protect account security.

i) Alert information from deployed security tools and security services shall be checked and handled in a timely manner. Based on other obtained security threat information, risk alerts, and similar information, the deployment and use of the agent shall be reviewed for security, and security risks shall be removed in a timely manner.

j) Security notices issued by the agent operator and by security companies shall be followed continuously. When a security incident affecting a specific version occurs, corresponding measures such as upgrade, downgrade, or version pinning shall be taken according to the security notice, in order to avoid long-term use of versions with known vulnerabilities or versions that have ceased to be maintained.

10 Decommissioning Stage

This chapter sets out security guidance for the decommissioning stage of the agent, including safely terminating operation, cleaning up data and permissions, and completing environment reclamation.

a) The main program of the agent, as well as all related services and background processes, shall be stopped, and it shall be confirmed that the agent has completely exited its running state.

b) Before cleaning up the environment, conversation records, configuration files, knowledge-base files, logs, and other data that still need to be retained shall be backed up for disaster recovery.

c) After completing the necessary data backup, corresponding environment-cleanup operations shall be performed according to the deployment environment:

1. Where deployed in a local environment, if the agent provider provides an official uninstaller, the official uninstaller should be used for removal, and after removal the relevant processes, ports, network connections, and other conditions should be checked to confirm that no residue remains. If there is no official uninstaller, the operating system should be reset.

2. Where deployed in a cloud environment, public access entry points related to the agent shall be closed, unused working files and data such as the agent’s knowledge bases, logs, long-term memory, plugins, and skill configurations shall be deleted, and related credentials and permissions such as API keys, service-account credentials, and third-party application authorizations shall be revoked at the same time.

3. Where deployed in a virtualized environment, the virtualized environment itself should be decommissioned directly if possible. If that virtualized environment cannot be decommissioned, environment cleanup should be carried out with reference to the relevant guidance in 1).

4. After decommissioning the agent, related services should be confirmed as terminated in a timely manner, subscriptions and auto-renewal should be canceled, and attention should be paid to whether the large-model interface is generating abnormal charges.

Appendix A (Informative) Security Checklist

Table 1 Security Checklist for Deployment and Use

Appendix B (Informative) Security Management Guidance

Where agents are used within an organization, it is recommended that management work be carried out around the following aspects.

a) Establish an internal management regime for agent use, and manage activities such as the deployment, use, and change of agents. The management regime shall include at least the following:

1. Requirements on prohibited behaviors. Specific behaviors prohibited during agent use should be clearly defined, for example, not connecting internal organizational data or system interfaces to agents that have not passed approval.

2. Requirements on usage boundaries. In accordance with the principle of minimum authorization, the scope of use, permission boundaries, and data-processing boundaries of agents should be clearly defined.

3. Requirements on approval procedures. The processes for application, assessment, approval, and filing before agent deployment should be clearly defined. The submitted application content should include the purpose of use, usage boundaries, skills to be configured, tools to be connected, security protection measures, and similar matters.

b) Establish an asset register for agents, carry out unified registration of approved agents, and update the register when changes occur. The asset register shall record at least the following information:

1. Agent information, including the agent name, purpose, and business ownership.

2. Deployment-location information, including host physical location, IP address, container name, or another uniquely identifiable agent identifier.

3. Developer information, including the deployer, deployment time, and responsible person.

4. Model information, including the name, version, and access method of the large model in use.

5. Tool and component information, including the list of installed skills, plugins, and other external tools.

6. Access-control information, including the accessible internal systems, interfaces, and data scope.

7. Authorization information, including permission configuration status.

8. Review-record information, including the time and result of the most recent security review.

c) Manage the activities of approved agents:

1. Keep logs of key activities, including execution of high-privilege commands, access to sensitive data, access to internal systems, transmission of external data, permission changes, tool calls, and abnormal operations.

2. Use artificial intelligence, rule analysis, or other technical means to conduct risk analysis of agent behavior patterns.

3. Conduct periodic analysis and review of logs to identify and handle security issues in a timely manner, and form records.

d) The ability to discover unapproved agents should be established through the following means, so that further handling such as network isolation, permission downgrading, and process blocking can be carried out:

1. Regularly scan typical agent service ports within the organizational intranet.

2. Analyze network-traffic logs to identify abnormal agent communication behavior.

3. Identify internal hosts communicating with known large-model API endpoints.

4. Search endpoints for processes related to the installation or operation of agents.

e) Conduct security education for employees within the organization to improve their awareness of agent security risks and secure use. Security education shall cover prompt-injection attacks, supply-chain security, credential leakage, data leakage, unauthorized access, liability for improper use, and similar topics.

Compared with Version 1 released on May 25, the new paper adds substantial detail on engineering implementation, measured quantitative data, and product evolution roadmaps on top of the original theoretical framework. It further refines the post-Moore scaling theory system centred on the time constant τ.

On engineering implementation, Version 2 provides a more in-depth explanation of the gear ratio concept in the core technology LogicFolding. When hybrid bonding pitch approaches the size of top-level metal interconnects, the 3D design space shifts from traditional "macro-block-level discrete optimisation" to "cell-level continuous optimisation." This makes globally optimal vertical logic partitioning possible and breaks the traditional limitation of 3D stacking to functional-block-level layering only.

Version 2 also adds a new table of measured mass-production data, explicitly listing voltage, frequency, normalised power consumption, area, and power density parameters for the Kirin 2026 and the benchmark Kirin 9030 Pro.

In May this year, at the 2026 International Symposium on Circuits and Systems, He Tingbo officially introduced the "Tao (τ) Law" — the first time China had proposed an industry development guiding principle in the global semiconductor sector.

The core of the Tao Law is "time miniaturisation." Rather than simply pursuing the extreme reduction of physical dimensions, it shifts toward the systematic reduction of the time constant (τ) — compressing signal transmission delay inside chips. The framework establishes a multi-level collaborative optimisation system spanning devices, circuits, chips, and systems. By 2031, high-end chips based on this law are projected to reach transistor density equivalent to the 1.4 nm process node.

If you scroll to the bottom of almost any Chinese website, you will find a small line of text that reads something like: “Beijing ICP Filing No. XXXXX.”（京ICP备XXXXX号）

That filing number traces its origins to a 25-year-old administrative regulation: the Measures for the Administration of Internet Information Services.

On September 25, 2000, then-Premier Zhu Rongji signed State Council Order No. 292, promulgating the Measures alongside China’s Telecommunications Regulations. At the time, China had only around 20 million internet users. The commercial internet was still in its infancy, and NetEase, Sohu, and Sina had only just listed on U.S. stock exchanges.

The regulation consisted of just 27 articles, but it established the basic regulatory architecture that would govern China’s internet for the next two decades. Commercial internet information services were subject to licensing, while non-commercial websites only needed to complete an ICP filing. A limited number of sectors—including news, publishing, education, healthcare, and pharmaceuticals—required prior administrative approval before they could operate online.

That framework has remained remarkably durable. Twenty-five years later, however, the regulation is undergoing what amounts to a complete rewrite, judging by the 3 July draft.

The new draft has expanded from 27 articles in 2000 to 94 articles, almost doubling in size compared with the 54-article draft published for consultation in 2021.

The graphic below provides a quick overview of the key revisions.

In fact, the CAC had already attempted a major overhaul in 2021, working jointly with the Ministry of Industry and Information Technology (MIIT) and the Ministry of Public Security. That draft introduced a wide range of new concepts—including real-name registration, platform responsibilities, log retention requirements, and enhanced content governance—but it was never formally adopted.

The latest revision reflects just how dramatically China’s internet landscape has changed over the past five years. Generative AI, recommendation algorithms, livestreaming, short-video platforms, and the platform economy have all expanded at a pace that the existing regulatory framework can no longer adequately address.

If the 2000 regulation was primarily about regulating Internet Service Providers (ICPs), the 2026 draft is fundamentally about regulating digital platforms and artificial intelligence.

One of the most visible changes concerns the allocation of regulatory authority.

When the regulation was first issued, the lead regulator was the then-Ministry of Information Industry—the predecessor of today’s MIIT—which oversaw market entry for internet services, while the Ministry of Public Security and state security authorities handled security-related matters.

The new draft reverses that hierarchy. The Cyberspace Administration of China is now explicitly identified as the primary authority, responsible for coordinating national cybersecurity policy and supervising online content nationwide. MIIT retains responsibility mainly for industrial matters such as market access, telecommunications resources, and market order. In other words, the institutional arrangement that has gradually emerged over the past decade—“the CAC leads, MIIT manages industry”—is now formally embedded in administrative law.

The traditional ICP licensing system is also being significantly restructured.

For more than twenty years, China’s internet has operated under a dual-track system: commercial services required an ICP license, while non-commercial services only needed an ICP filing. Whether a website charged users largely determined which regime applied.

The new draft largely removes this distinction by introducing a unified approval framework. It also provides that ICP approvals should be revoked if services fail to commence operations within two years or no longer satisfy the statutory requirements. Meanwhile, the list of sectors requiring prior approval has expanded well beyond the original nine categories, now covering areas such as audiovisual services, religion, and financial services.

The draft also significantly tightens the rules governing foreign-invested and cross-border internet services. Articles 91, 92, and 38 point in the same direction. Whether a service is foreign-funded, operated from overseas, or capable of shaping public opinion inside China, regulators are making it clear that substance matters more than corporate form.

Article 91 does not create a new foreign investment regime. Instead, it reaffirms that internet services remain subject to China’s foreign investment restrictions, including the Negative List. For companies using Variable Interest Entity (VIE) structures, this could be significant. For years, many internet companies have relied on contractual arrangements rather than direct foreign ownership to comply with investment rules. The new draft suggests regulators may increasingly look beyond legal structures to examine who actually controls the business, who receives the economic benefits, and what the platform actually does. That is especially relevant for businesses such as search, content distribution, online communities, recommendation algorithms, and generative AI, where the regulatory sensitivity is much higher.

Article 92 addresses another longstanding question: what happens if the company is overseas but the service is clearly aimed at users in China? The answer is increasingly straightforward. Being incorporated abroad or hosting servers overseas will no longer, by itself, shield a platform from Chinese regulation. If a service targets Chinese users or has a substantial impact inside China, it may still fall within China’s internet governance framework. For foreign social media platforms, search engines, and AI services, the issue is no longer just market access—it is whether they can comply with China’s rules on content moderation, algorithms, data governance, and platform responsibility.

Article 38 pushes the logic even further. It focuses on who ultimately controls services with significant public communication capabilities. If a platform can distribute news, aggregate information, or shape public opinion at scale, regulators are unlikely to treat it as just another commercial platform. For VIE companies, separating news-related businesses on paper may no longer be enough. For foreign platforms, businesses involving news or public discourse are likely to face much closer scrutiny over ownership and control.

Taken together, these provisions do more than tighten foreign investment rules. They redefine how China regulates cross-border internet services. Who provides the service, where it operates, and what it actually does all matter. For VIE companies, the regulatory flexibility created by complex ownership structures is likely to shrink. For foreign social media platforms, search engines, and AI providers, the biggest challenge may not be technology, but fitting into China’s increasingly unified internet governance framework. The draft is not targeting one particular business model—it is narrowing the regulatory grey zone that many companies have relied on for years.

In the new draft, there are also two entirely new chapters.

The first introduces a dedicated regulatory framework for platform information services.

For the first time, an administrative regulation defines what constitutes a “large internet platform.” The threshold is set at either more than 50 million registered users or more than 10 million monthly active users. By that definition, virtually every major Chinese internet platform—including WeChat, Douyin, Weibo, Bilibili, Kuaishou, Xiaohongshu, Pinduoduo, JD.com, Meituan, Didi, and Doubao—falls within its scope.

These platforms would face a series of additional obligations, including establishing comprehensive compliance systems, dedicated compliance departments, independent external oversight bodies, 24-hour complaint-handling mechanisms, annual social responsibility reports, reporting major operational changes, and complying with algorithm-related competition rules. Platform governance, previously governed largely through ministerial rules, is now being elevated to the level of an administrative regulation.

The second new chapter focuses on intelligent information services, marking the first time that generative AI has been comprehensively regulated through an administrative regulation rather than lower-level rules.

The draft introduces requirements covering training data provenance, mandatory labeling of AI-generated content, prohibitions on tampering with those labels, algorithm registration, and the option for users to disable personalized recommendation systems. It also marks the first appearance of the term “AI agents” in an administrative regulation.

Another notable provision addresses algorithmic management of gig workers. The draft requires algorithms affecting delivery riders, ride-hailing drivers, and other platform workers to be transparent, fair, and reasonable. Significant changes to algorithms governing recruitment, dispatching, or performance evaluation should involve consultation with affected workers.

Enforcement is also becoming substantially tougher. Historically, internet-related violations primarily resulted in penalties against companies.

Notably, the new draft introduces a dual-liability regime, under which both companies and responsible individuals may be sanctioned. Individuals directly responsible for violations face fines of up to RMB 1 million and, in serious cases, may be barred from serving as directors, supervisors, senior executives, or information security officers of internet companies. Corporate fines also increase significantly, with the maximum penalty rising from RMB 1 million to RMB 10 million.

This approach closely resembles developments in China’s financial and securities regulation, where accountability has increasingly shifted from institutions alone to the executives responsible for their conduct.

Taken together, the draft reveals several broader trends. To make the changes easier to digest, I've summarized the key implications in the chart below.

First, it consolidates obligations that were previously scattered across the Cybersecurity Law, Data Security Law, Personal Information Protection Law, Anti-Telecom and Online Fraud Law, and numerous ministerial regulations into a single administrative framework.

Second, it formally codifies the CAC’s central coordinating role in internet governance.

Third, it brings fast-growing technologies—including generative AI, deepfakes, recommendation algorithms, AI agents, livestreaming, and the platform economy—under a unified administrative regulatory framework.

Fourth, it is increasingly willing to look through corporate structures and regulate any foreign-linked internet service involved in information dissemination or public opinion inside China based on how it actually operates, not how it is legally structured.

Finally, it continues China’s broader shift toward institutionalizing platform governance while extending personal accountability through dual-liability provisions and executive disqualification mechanisms.

Below are the Chinese translations of the new draft and the CAC’s explanatory note:

Measures for the Administration of Internet Information Services

Draft Revision for Public Comment

Chapter I General Provisions

Article 1 These Measures are formulated in accordance with the Cybersecurity Law of the People’s Republic of China and other laws, for the purposes of promoting the healthy and orderly development of internet information services, protecting the lawful rights and interests of citizens, legal persons, and other organizations, and safeguarding national security and the public interest.

Article 2 These Measures shall apply to the provision of internet information services within the territory of the People’s Republic of China, as well as to the supervision and administration of internet information services.

Article 3 The provision of internet information services, as well as the supervision and administration of internet information services, shall uphold the leadership of the Communist Party of China, take the core socialist values as guidance, and follow the principles of being people-centered, classified management, coordinated advancement, and innovation-driven development, so as to promote the formation of a positive, healthy, uplifting and good-oriented online ecosystem and create a clean and upright cyberspace.

Article 4 The national cyberspace administration department shall be responsible for coordinating national cybersecurity work and related supervision and administration, and shall conduct supervision, administration, and law enforcement over internet information content nationwide.

The telecommunications department under the State Council shall, in accordance with its duties, be responsible for the administration of the national internet industry, and shall supervise and administer market access, market order, network resources, cybersecurity, and other matters relating to internet information services.

The public security department under the State Council shall, in accordance with its duties, be responsible for cybersecurity protection, supervision, and administration, maintaining public order and information security on the internet, and preventing and punishing illegal and criminal online activities.

State security organs shall, in accordance with their duties, be responsible for combating illegal and criminal activities that use the internet to endanger national security.

Other relevant departments under the State Council shall supervise and administer internet information services within the scope of their respective duties.

Local responsibilities for the supervision and administration of internet information services shall be determined in accordance with relevant national provisions.

Article 5 The State protects the rights of citizens, legal persons, and other organizations to use internet information services in accordance with the law, promotes the popularization of internet applications, and improves the level of internet information services.

The State encourages basic research and technological innovation related to internet information services, promotes the deep integration of the internet with economic and social development, encourages the building of an open, shared, mutually beneficial, and win-win industry development ecosystem, and enhances the innovative vitality of internet information services.

Those providing internet information services shall comply with laws and administrative regulations, perform cybersecurity obligations, assume social responsibilities, respect social morality and ethics, observe business ethics, act in good faith, and accept supervision by the government and society.

Article 6 The State encourages industry organizations in the internet information services sector to strengthen industry self-discipline, establish and improve industry self-discipline systems and industry standards, guide internet information service providers in formulating and improving service rules, urge internet information service providers to provide services in accordance with the law and accept social supervision, and promote the healthy and orderly development of the industry.

Article 7 The State shall take measures to monitor, prevent, and handle illegal and criminal activities conducted through internet information services inside or outside the territory of the People’s Republic of China that harm national security, the public interest, or the lawful rights and interests of citizens and organizations of the People’s Republic of China.

Chapter II Establishment

Article 8 Anyone providing internet information services within the territory of the People’s Republic of China shall obtain approval from the telecommunications department in accordance with the law. Without approval, no entity or individual may provide internet information services.

Article 9 Where the provision of internet information services constitutes the operation of telecommunications business, an application shall be made to the telecommunications department to obtain the relevant telecommunications business operating license.

The telecommunications department shall, in accordance with the requirements prescribed by relevant laws and administrative regulations, make a decision to approve or not approve the application.

Where an internet information service provider that has obtained the relevant telecommunications business operating license has not carried out the relevant telecommunications business for two consecutive years, or no longer meets the conditions for the telecommunications business operating license, the telecommunications department that made the licensing decision shall cancel its relevant telecommunications business operating license.

The telecommunications department shall establish systems for classified administration of telecommunications business and market monitoring. Relevant internet information service providers shall submit information on network resources, telecommunications business operations, and other matters as required.

Article 10 Anyone providing internet information services shall use network resources that meet the requirements of the telecommunications department, and shall have cybersecurity and information security management systems and technical safeguard measures that comply with laws, administrative regulations, and national provisions. The provision of internet information services shall comply with the requirements of mandatory national standards for network products, services, and related matters.

Internet websites and applications are encouraged to support and give priority to connections and access using Internet Protocol Version 6.

Article 11 Where the provision of internet information services in areas such as news, culture, publishing, audiovisual programs, education, religion, and finance requires approval from relevant competent authorities in accordance with laws, administrative regulations, and relevant national decisions, such approval shall be obtained before completing approval procedures or applying for a telecommunications business operating license.

Article 12 Personnel engaged in internet information services shall adhere to the correct political orientation, public opinion guidance, and value orientation, comply with laws and administrative regulations, and possess corresponding professional ethics and professional competence. Personnel engaged in internet news information services shall obtain the corresponding qualifications and receive the corresponding training and assessment. Specific measures for such qualifications shall be formulated by the national press and publication authority and the national cyberspace administration department in accordance with their respective duties.

Internet information service providers shall, in accordance with relevant national provisions, strengthen the management and training of personnel engaged in internet information services, especially personnel responsible for information security review.

Article 13 Where an internet information service provider no longer provides internet information services, it shall promptly complete the cancellation procedures for the relevant license or approval.

Where an internet information service provider undergoes a change in the sponsor-related licensing or approval matters due to merger, acquisition, or other reasons, it shall complete change procedures with the original licensing or approval authority.

Chapter III Operations

Section 1 General Provisions

Article 14

Where an internet information service provider applies for internet access services, it shall provide the internet access service provider with its approval number and the corresponding license or approval documents. The internet access service provider shall promptly verify such information and shall not provide services to any internet information service provider that has not obtained the required approval number and supporting documentation.

The establishment of root domain name servers, root domain name server operating institutions, domain name registration management institutions, domain name registration service institutions, or the provision of recursive domain name resolution services shall obtain the approval of the telecommunications department in accordance with the law.

Internet domain names shall not contain any content prohibited by laws or administrative regulations. Domain names used by internet information service providers shall be lawfully registered in their own names. Where a domain name is changed, transferred, or expires, the provider shall promptly complete the corresponding amendment or cancellation procedures for its internet information service approval. Internet information service providers shall not use domain names to carry out unlawful activities.

Article 15

The State implements a filing system for the allocation and use of Internet Protocol (IP) addresses. IP address allocation institutions shall promptly file relevant IP address information with the telecommunications department.

Before providing internet access services, internet access service providers shall verify the relevant IP address filing information and shall not provide services for IP addresses that have not been registered under a real identity or that contain false registration information.

Article 16

Any person providing internet information services within the territory of the People’s Republic of China shall, within 30 days from the date on which the network is officially connected, complete the required filing procedures with the public security authority designated by the people’s government at the provincial level where the provider is located. Internet access service providers shall provide the necessary assistance.

Internet information service providers shall prominently display their public security filing number in an easily accessible location.

Article 17

When providing internet information services, providers shall prominently display, in an easily accessible location, the license number or approval number obtained pursuant to Articles 8, 9, and 11 of these Measures.

Internet information service providers shall provide services only within the scope of the relevant license or approval and shall not provide services beyond the approved scope.

Article 18

Internet information service providers shall establish and improve systems governing user registration, account management, information publication, public information inspection, emergency response, and content governance. They shall enter into service agreements with users in accordance with the law and possess professional personnel and technical capabilities commensurate with the scale and nature of their services.

Internet information service providers shall strengthen the management of user account information and promptly conduct dynamic verification of account information. Where account information is found to contain information that violates laws, administrative regulations, or relevant national provisions, the provider shall refuse account registration or suspend the provision of the relevant services.

Where users engage in activities through internet information services that require specific qualifications under laws, administrative regulations, or relevant national provisions, the users shall provide proof of such qualifications to the internet information service provider. The provider shall verify such documentation and shall not provide the corresponding services to users lacking the required qualifications.

Article 19

Providers of internet access services, internet information services, domain name registration services, domain name resolution services, and similar services shall, when entering into service agreements or confirming the provision of services, require users to provide their true identity information in accordance with the law. Where users fail to provide true identity information, the relevant services shall not be provided.

Such service providers shall authenticate users’ true identity information through the National Online Identity Authentication Public Service, identity document numbers, Unified Social Credit Codes, or other legally prescribed means, and shall retain the relevant information. Verified identity information shall be retained throughout the period during which the service is provided and shall continue to be retained after termination of the service in accordance with the law.

Internet information service providers shall verify the true identity information of minor users through necessary means in accordance with the law.

Article 20

No organization or individual may evade real identity authentication by engaging in any of the following conduct:

1. using false identity information or impersonating another person’s identity to provide or use internet information services;

2. illegally obtaining or using internet accounts or internet resources registered by others; or

3. providing technical support, assistance, or other help to others for the purpose of evading real identity authentication.

Article 21

Internet information service providers shall establish a system for dynamic verification of internet account information and shall periodically verify existing account information. For abnormal internet accounts, providers may conduct dynamic verification through identity document numbers, the National Online Identity Authentication Public Service, or other lawful means.

Where an account is found to be inconsistent with the user’s verified identity information, or has remained inactive for more than six months, the provider shall, in accordance with the law and the service agreement, issue appropriate reminders and may suspend, restrict, or prohibit information publication or related functions, and, where necessary, freeze, cancel, or permanently close the account or prohibit re-registration.

Where the user of a mobile telephone number changes, the internet information service provider shall support the removal of the binding relationship between the mobile telephone number and the former user’s internet account.

Article 22

Providers of internet access services, internet information services, domain name registration services, domain name resolution services, and similar services shall adopt technical and other necessary measures to prevent, detect, and stop the use of their services for unlawful or criminal activities.

Where such providers discover cybercrime or other unlawful online activities, they shall preserve the relevant records and report them to the public security authorities and other competent authorities.

Article 23

The State encourages internet information service providers and users to produce, reproduce, publish, disseminate, and transmit information that promotes the core socialist values, advanced socialist culture, revolutionary culture, fine traditional Chinese culture, and socialist rule-of-law culture; reflects the great efforts of the Chinese people and the positive spirit of the Chinese nation; showcases achievements in economic and social development and the development of the rule of law in China; and promotes national reunification and ethnic unity.

Article 24

Internet information service providers and users shall not produce, reproduce, publish, disseminate, or transmit information containing any of the following:

1. information opposing the fundamental principles established by the Constitution; endangering national security, national honor, or national interests; divulging State secrets; inciting the subversion of State power; overthrowing the socialist system; inciting the splitting of the country; or undermining national unity;

2. information advocating terrorism or extremism, or inciting terrorist or extremist activities;

3. information promoting ethnic hatred or ethnic discrimination, undermining ethnic unity, harming public morality or China’s fine traditional culture, undermining the State’s religious policies, using religion to obstruct the implementation of the State’s judicial, educational, or social administration systems, deceiving or inciting others to engage in unlawful religious activities, or promoting cults or superstition;

4. information inciting unlawful assemblies, associations, marches, demonstrations, or other activities that undermine social stability;

5. information that distorts, vilifies, desecrates, or denies the deeds and spirit of heroes and martyrs; infringes upon the names, portraits, reputations, or honors of heroes and martyrs through insult, defamation, or other means; disparages or damages the honor of military personnel or veterans; insults or defames their reputation; intentionally destroys or desecrates honorary symbols of military personnel or veterans; or otherwise harms national defense or military development;

6. information disseminating obscenity, pornography, violence, gambling, homicide, terrorism, or content that incites crime, teaches criminal methods, or involves the manufacture or trading of prohibited or controlled items;

7. false or misleading information that disrupts economic or social order;

8. information insulting or defaming others or infringing upon another person’s reputation, honor, privacy, personal information rights and interests, intellectual property rights, or other lawful rights and interests; or

9. other information prohibited by laws or administrative regulations.

Article 25

Internet information service providers shall adopt measures to prevent and resist the production, reproduction, publication, dissemination, or transmission of undesirable information involving major national policies, emergencies, major criminal cases, or content that may adversely affect the physical or mental health of minors.

Internet information service providers shall not prominently display such undesirable information in products or services, or at locations or stages that are likely to attract users’ attention.

The State encourages cooperation between internet information service providers and news and publishing institutions to improve content governance.

Article 26

No organization or individual shall engage in any of the following conduct that disrupts the order of internet information services:

1. publishing, deleting, blocking, disconnecting links to, replacing, suppressing, or algorithmically recommending information, or providing such services for others;

2. fraudulently or in bulk registering internet accounts, or illegally stockpiling or trading internet accounts;

3. manipulating or using multiple internet accounts to mass-publish information prohibited by laws or administrative regulations, or undesirable information; engaging in fraudulent clicks, voting, rankings, comments, transactions, or reviews; fabricating or hijacking internet traffic; creating false public opinion hotspots; or manipulating rankings, trending lists, or other key online mechanisms;

4. unlawfully bypassing or circumventing technical measures adopted by relevant State authorities; or

5. other conduct prohibited by laws or administrative regulations.

Internet information service providers shall adopt measures such as abnormal data monitoring and manual review to prevent and detect the foregoing conduct.

Article 27

Where an internet information service provider discovers information prohibited under Article 24, undesirable information referred to in Article 25, or conduct prohibited under Article 26, it shall immediately stop transmission, take measures including deletion, blocking, disconnection of links, restriction of account functions or monetization rights, or account closure to prevent further dissemination, preserve relevant records, and report the matter to the cyberspace administration and other competent authorities.

Providers of electronic information transmission services and application download services that become aware that their users have produced, reproduced, published, disseminated, or transmitted prohibited or undesirable information, or engaged in conduct prohibited under Article 26, shall likewise immediately stop transmission, adopt deletion, blocking, link disconnection, or other necessary measures, preserve relevant records, and report to the cyberspace administration and other competent authorities.

Article 28

Where an internet access service provider discovers circumstances involving information prohibited under Article 24 or undesirable information referred to in Article 25, it shall immediately stop transmission, adopt appropriate measures including removal of the relevant information to prevent further dissemination, preserve relevant records, and report the matter to the cyberspace administration, telecommunications authority, public security authority, and other competent authorities.

Article 29

Where the competent authorities, in performing their statutory responsibilities for online information security supervision, discover information prohibited under Article 24, undesirable information referred to in Article 25, or conduct prohibited under Article 26, they shall, within the scope of their respective duties, require internet information service providers and internet access service providers to take the measures specified in Articles 27 and 28. Internet information service providers and internet access service providers shall cooperate accordingly.

Where organizations or individuals outside the territory of the People’s Republic of China produce, reproduce, publish, disseminate, or transmit information prohibited under Article 24 that harms China’s national security or public interests, the relevant State authorities may take necessary measures against such organizations or individuals in accordance with applicable laws and regulations.

Article 30

Where a person knows that another person is engaging in any of the following acts in violation of these Measures, no organization or individual shall provide data, technology, software, tools, programs, advertising, services, payment settlement, or any other support or assistance:

1. producing, reproducing, publishing, disseminating, or transmitting information prohibited under Article 24 or undesirable information referred to in Article 25;

2. engaging in conduct prohibited under Article 26 that disrupts the order of internet information services; or

3. assisting others in obtaining or disseminating information that has been lawfully blocked.

Article 31

Internet information service providers shall establish and improve rumor-refutation mechanisms. Where false information is produced, reproduced, published, disseminated, or transmitted, providers shall promptly adopt rumor-refuting measures, preserve relevant records, and report to the cyberspace administration and other competent authorities.

Where the conduct is suspected of constituting a criminal offense, the provider shall promptly report the matter to the public security authorities.

Article 32

Internet information service providers shall not engage in any of the following acts that infringe upon the lawful rights and interests of other internet information service providers:

1. without justifiable reason, interfering with the downloading, installation, operation, upgrading, or uninstallation of services or products lawfully provided by other internet information service providers on users’ terminal devices;

2. without justifiable reason, intercepting, blocking, or rendering incompatible services or products lawfully provided by other internet information service providers, or blocking hyperlinks, application redirects, or similar functions;

3. without the consent of another internet information service provider, inducing users to access the provider’s own services or those of a designated third party by inserting links, embedding content, creating redirects, or using misleading information, drop-down menus, search suggestions, interface designs, or operation buttons within services or products lawfully provided by another provider;

4. without justifiable reason, modifying, disabling, or uninstalling services or products lawfully provided by another internet information service provider, or deceiving, misleading, or compelling users to do so; or

5. other acts that violate laws, administrative regulations, or relevant national provisions and infringe upon the lawful rights and interests of other internet information service providers.

Article 33

Internet information service providers shall truthfully and comprehensively inform users of service information and applicable charges, respect users’ right to choose services independently, and shall not add, subscribe users to, or modify services without authorization, nor provide services through misleading practices, deception, or bundled offerings.

Internet information service providers shall strengthen service quality management and conduct service quality monitoring and evaluation. Relevant authorities, including the telecommunications, cyberspace administration, and market regulation authorities, may supervise and encourage providers to improve service quality through public service quality evaluations and other means.

Article 34

Internet information service providers and internet access service providers shall establish systems for public complaints, reporting, and user appeals, publicly disclose the channels for complaints, reporting, and appeals, provide human customer service resources commensurate with the scale of their business, promptly handle complaints, reports, and appeals, and communicate the handling results to users.

Article 35

The State encourages and supports internet information service providers in developing and providing services such as youth modes and dedicated youth sections that are suitable for the physical and mental development of minors and facilitate their access to beneficial internet information services.

Internet information service providers shall fulfill their statutory obligations for the online protection of minors, adopt necessary technical and management measures to prevent cyberbullying targeting minors, and prevent minors from becoming excessively dependent on or addicted to internet information services.

Article 36

Internet information service providers shall protect the lawful rights and interests of elderly persons and persons with disabilities in accordance with the law, fully consider their needs, and provide age-friendly services, technical accessibility support, and barrier-free services in accordance with relevant national provisions and standards.

Article 37

Internet information service providers that offer services possessing the attributes of public opinion or the capacity for social mobilization shall conduct security assessments in accordance with relevant national provisions.

The competent authorities shall, in accordance with relevant national provisions, conduct security assessments of internet information services possessing the attributes of public opinion or the capacity for social mobilization.

Article 38

Eligible internet information service providers operating in the news, publishing, and media sectors shall implement the Special Management Share (Special Management Stock) system in accordance with relevant national provisions.

Article 39

Internet information service providers shall record and retain information published through their services, user registration information, records of user-generated content, and other network log information for no less than six months.

Where internet access service resources are shared with others through network proxy services, network address translation (NAT), or similar technologies, providers shall also record and retain address translation logs and other log information sufficient to identify users.

Article 40

The State encourages internet information service providers to give priority to the use of open-source operating systems when developing applications.

Interconnection and interoperability among internet information service providers shall comply with the requirements of the cyberspace administration, telecommunications authorities, and other competent authorities. Such interconnection shall not infringe upon the lawful rights and interests of third parties, nor may it be used to publish or disseminate information prohibited by laws or administrative regulations.

Chapter III Operations

Section 2 Platform Information Services

Article 41

Internet information service providers that offer internet information dissemination platform services (hereinafter “Platform Information Service Providers”) shall continuously publish their platform rules, or a clearly identifiable link to such rules, in a prominent location on the homepage of their websites or applications, ensuring that relevant parties can conveniently access, review, and download them.

The platform rules shall prominently require users of internet information services to comply with laws, administrative regulations, and relevant national provisions.

Where platform rules are newly introduced or amended, Platform Information Service Providers shall publicly disclose the proposed changes and adequately solicit opinions from relevant stakeholders.

Where platform rules violate laws, administrative regulations, or relevant national provisions, the competent authorities shall, within the scope of their respective responsibilities, order the provider to make appropriate corrections.

Article 42

Platform Information Service Providers shall establish a credit record system for internet accounts, reasonably determine account credit ratings based on account usage, provide differentiated services accordingly, and periodically report relevant information to the cyberspace administration and other competent authorities.

Article 43

Where an internet public account applies to provide internet news information services, online publishing services, or other internet information services that require administrative approval by law, or applies to produce and disseminate information in areas such as finance, education, healthcare, legal affairs, military affairs, religion, human resources, or funeral services, the Platform Information Service Provider shall require the operator of the public account to provide information regarding relevant professional qualifications, professional background, and licenses or permits required by laws and administrative regulations.

The provider shall verify such information and apply a special identifier to the account.

Article 44

Platform Information Service Providers shall display, on the profile page of each internet public account, information including:

• the operating entity;

• the Unified Social Credit Code;

• valid contact information;

• the geographic location associated with the account’s IP address; and

• the affiliated multi-channel internet content distribution service organization, where applicable.

Article 45

Operators of internet public accounts with significant influence shall not produce, reproduce, publish, disseminate, or transmit information prohibited under Article 24 or undesirable information referred to in Article 25.

Where such operators violate the preceding paragraph, Platform Information Service Providers shall, in accordance with law and contractual agreements, adopt measures including warnings, rectification orders, suspension of updates, restrictions on account functions, suspension of monetization privileges, account closure, and prohibition of re-registration, promptly remove the unlawful or undesirable content, preserve relevant records, and report the matter to the cyberspace administration and other competent authorities.

Specific measures for determining which internet public accounts possess significant influence shall be formulated by the national cyberspace administration department in conjunction with the national press and publication authority and the competent departments of culture and tourism, radio and television, public security, and other relevant authorities under the State Council.

Article 46

Platform Information Service Providers shall, in accordance with relevant national provisions, file with the competent authorities information relating to internet public accounts with significant influence, including the account name, operator, and affiliated multi-channel internet content distribution service organization.

Providers of internet livestreaming platform services shall complete the required filing with the provincial-level cyberspace administration department within 30 days after commencing operations. The cyberspace administration department shall promptly share the filing information with relevant authorities.

Article 47

Platform Information Service Providers shall establish and improve databases of characteristics of cyberbullying information and databases of representative cyberbullying cases, and strengthen the identification and monitoring of cyberbullying information through a combination of technical measures and manual review.

They shall also establish and improve early warning models for cyberbullying information in order to identify and warn of cyberbullying risks in a timely manner.

Platform Information Service Providers shall guide users toward civil and rational online interaction. Where cyberbullying risks are identified, providers shall promptly adopt measures including dynamic verification of users’ real identities, pop-up reminders, legal warnings, and traffic restrictions for abnormal accounts.

Where the volume of views, comments, reports, or similar indicators relating to the relevant information increases significantly, providers shall promptly report the matter to the competent authorities.

Platform Information Service Providers shall also provide users with tools to protect themselves against cyberbullying, including options to block designated users, disable reposting or commenting, refuse all private messages, and rapidly preserve evidence relating to cyberbullying content.

Article 48

Platform Information Service Providers shall perform their obligations in accordance with these Measures and other applicable laws and administrative regulations, and shall not engage in conduct prohibited under Article 26 that disrupts the order of internet information services or circumvents regulatory oversight.

Platform Information Service Providers shall also periodically conduct self-assessments regarding their fulfillment of internet information service management obligations and compliance with laws and administrative regulations.

Article 49

Platform Information Service Providers shall not abuse their market influence, platform rules, user agreements, or rely on the pretext of fulfilling statutory obligations to disrupt the order of internet information services or evade regulatory oversight.

Article 50

The State shall improve the market access regime for the internet livestreaming industry and strengthen full lifecycle management of livestreaming accounts.

Providers of internet livestreaming platform services shall establish dynamic real-identity verification mechanisms for livestreaming accounts, implement classified and tiered management based on factors such as livestreaming content and audience size, and strengthen real-time management of interactive features including comments and bullet chats (danmu).

Article 51

Providers of application distribution platform services shall complete the required filing in accordance with laws, administrative regulations, and relevant national provisions before providing application distribution services.

Application distribution platform providers shall strengthen the management of applications available on their platforms. They shall give greater visibility and promotional support to applications that actively disseminate the content encouraged under Article 23 of these Measures and voluntarily maintain a healthy online environment.

Application distribution platform providers shall register and verify the qualifications, principal functions, and basic information of application developers, strengthen ongoing inspections of applications, and cease providing distribution services to applications that commit serious violations of laws or regulations.

Neither application distribution platform providers nor application service providers shall induce users to download applications through false advertising, bundled downloads, manipulation of rankings, fabricated traffic, fake transactions or reviews, or by publishing or disseminating information prohibited by laws or administrative regulations or other undesirable information.

Article 52

Providers of multi-channel internet content distribution services shall register as business entities in accordance with the law, appoint a person responsible for content management, establish a content review team commensurate with the scope and scale of their business, and formulate and publicly disclose rules governing content management.

Platform Information Service Providers shall strengthen the management of such organizations operating on their platforms, verify them in accordance with laws, administrative regulations, and relevant national provisions, and complete the required filing procedures in accordance with national requirements.

Article 53

Where a multi-channel internet content distribution service organization provides planning, production, publishing, dissemination, marketing, promotion, agency, or other services relating to internet content for operators of internet public accounts, it shall clearly specify content management responsibilities in the service agreement and perform its information security management obligations in accordance with the law and the agreement.

Where such an organization fails to fulfill its management obligations with respect to contracted public accounts that violate laws, administrative regulations, or these Measures, it shall bear the corresponding legal responsibilities in accordance with the law.

A multi-channel internet content distribution service organization shall not directly engage in, organize, incite, encourage, commission, or assist contracted public accounts in violating these Measures or other applicable laws and administrative regulations. It shall, in accordance with the law and contractual arrangements, adopt measures such as warnings, rectification orders, suspension of monetization privileges, termination of agreements, and notification to the relevant platform.

Where either the organization or its contracted public accounts violate these Measures or other applicable laws and administrative regulations, the Platform Information Service Provider shall, in accordance with the law and contractual arrangements, adopt measures including warnings, rectification orders, suspension of monetization privileges, restriction of services, removal from the platform, blacklisting, and other appropriate actions, and shall report the matter to the competent authorities in accordance with relevant national provisions.

Article 54

Large Platform Information Service Providers shall fulfill the following obligations:

1. establish a comprehensive compliance management system and risk prevention mechanisms for internet information services; establish a dedicated compliance department and appoint a compliance officer; conduct regular security assessments; and publicly disclose the operation of the relevant mechanisms and measures;

2. formulate and publish dedicated platform rules specifying the protection of the lawful rights and interests of providers of products or services on the platform, as well as workers directly involved in platform information services, together with the available remedies;

3. establish efficient mechanisms for handling public complaints, reports, and user appeals relating to internet information content, and process complaints, reports, and appeals concerning unlawful or undesirable information within 24 hours;

4. publish periodic social responsibility reports disclosing matters including internet information security, personal information protection, and the protection of minors online, and accept public supervision; and

5. promptly report to the competent local cyberspace administration department any changes to their principal business, major shareholders, actual controllers, or any major emergency incidents.

Large Platform Information Service Providers shall not use data, algorithms, technology, platform rules, technical measures, or user agreements to infringe upon the lawful rights and interests of other business operators, consumers, or other market participants.

Chapter III Operations

Section 3 Intelligent Information Services

Article 55

The State adheres to the principle of giving equal importance to development and security, and combining the promotion of innovation with law-based governance. It encourages the innovative application of artificial intelligence technologies in internet information services, advances the development of standards for algorithms, supports the establishment of high-quality datasets for internet information services, promotes the development, training, and application of AI models related to internet information services, and encourages intelligent information services to develop in a positive and beneficial direction.

Article 56

Providers of internet information services using artificial intelligence technologies (hereinafter “Intelligent Information Service Providers”) shall follow the principles of fairness, openness, transparency, and good faith. In accordance with relevant national provisions, they shall publicly disclose information concerning the basic principles, purposes, operating mechanisms, and sources of training data for the relevant technologies.

Article 57

Intelligent Information Service Providers offering services that possess the attributes of public opinion or the capacity for social mobilization shall conduct security assessments in accordance with relevant national provisions and complete the required algorithm filing, amendment, and deregistration procedures.

Article 58

Where an application service provider invokes system permissions of intelligent terminal devices to provide intelligent information services, the telecommunications and cyberspace administration authorities shall, within the scope of their respective responsibilities, organize compatibility testing between the application and the intelligent terminal.

Article 59

Where Intelligent Information Service Providers conduct data processing activities such as pre-training or post-training optimization, they shall use data and foundation models obtained from lawful sources, adopt effective measures to improve the quality of training data, and shall not infringe upon the intellectual property rights lawfully enjoyed by others.

Where corpus data is used for model training, such training shall comply with laws, administrative regulations, relevant national provisions, and the requirements of the core socialist values. Providers of internet news information services shall use as training corpora only news information published by entities approved under relevant national provisions.

Article 60

Intelligent Information Service Providers shall adopt both technical and manual measures to strengthen the management of AI-generated and synthetic content.

Providers offering AI-generated content services shall label AI-generated and synthetic content in accordance with laws, administrative regulations, relevant national provisions, and mandatory national standards.

No organization or individual shall maliciously remove, alter, forge, or conceal such labels.

Article 61

Intelligent Information Service Providers offering algorithm recommendation services shall, in accordance with relevant national provisions, establish mechanisms for human intervention and user choice, provide users with options that are not based on personal characteristics or offer convenient means to refuse personalized recommendations, and shall not compel users to use intelligent information services.

Providers offering algorithm recommendation services shall formulate traffic allocation rules and algorithm standards that promote positive values, file such rules with the relevant authorities in accordance with national requirements, and publicly disclose the relevant information.

Providers offering algorithm recommendation services shall regulate traffic allocation and promotional mechanisms, strengthen governance of recommendation interfaces, and actively present content consistent with mainstream values in prominent areas such as homepages, trending lists, featured content, rankings, and pop-up windows. They shall not use information prohibited by laws or administrative regulations, or undesirable information, as keywords in user interest profiles or user tags for recommendation purposes. Nor shall they use algorithms to suppress information, engage in excessive recommendation, manipulate rankings or search results, control trending topics or featured content, interfere with information presentation, disrupt online public opinion, or circumvent regulatory oversight.

Providers offering algorithm recommendation services shall configure recommendation models that comply with the information content requirements of these Measures and shall not use recommendation technologies to produce, reproduce, publish, disseminate, or transmit information prohibited under Article 24 or undesirable information referred to in Article 25.

Article 62

Where Intelligent Information Service Providers provide work scheduling or similar services to workers, they shall collect workers’ data in recruitment, dispatching, performance evaluation, and other employment-related processes in accordance with the principles of legality, legitimacy, and necessity. They shall ensure that algorithms governing order allocation, remuneration, working hours, and similar matters are transparent, fair, and reasonable, and shall protect workers’ lawful rights and interests, including equal employment opportunities, remuneration, and rest periods.

Where major algorithmic rules are introduced or modified, providers shall publicly solicit opinions from workers, fully consider the views of relevant stakeholders, assess the impact on workers’ lawful rights and interests, publicly disclose the changes in a prominent manner before implementation, and shall not infringe upon workers’ lawful rights and interests.

Article 63

Providers offering AI Agent services shall use AI Agents that comply with relevant national requirements and establish corresponding security management systems and technical control capabilities.

Providers engaged in AI Agent distribution services shall strengthen the security management of AI Agents in accordance with relevant national provisions and establish sound AI Agent security management systems.

Providers offering AI Agent services or AI Agent distribution services shall not use AI Agents to engage in activities that endanger national security, harm the public interest, or otherwise violate laws or administrative regulations.

Article 64

Where an Intelligent Information Service Provider discovers that its intelligent information services present significant security risks, generate or synthesize information prohibited under Article 24 or undesirable information referred to in Article 25, or involve conduct prohibited under Article 26, it shall immediately adopt measures including suspending content generation, stopping transmission, removing relevant content, suspending services, or disabling relevant functions. It shall also take corrective measures such as further model optimization and training, and report the matter to the cyberspace administration and other competent authorities.

Chapter IV Supervision and Inspection

Article 65

The cyberspace administration, telecommunications, public security, and other competent authorities shall, within the scope of their respective responsibilities, supervise and inspect internet information services and promptly investigate and deal with violations of these Measures.

When the cyberspace administration, telecommunications, public security, or other competent authorities perform supervisory and inspection duties in accordance with the law, internet information service providers and internet access service providers shall cooperate and shall not refuse or obstruct such supervision or inspection.

Article 66

The cyberspace administration, telecommunications, public security, and other competent authorities shall strengthen coordination and information sharing, establish mechanisms for sharing supervisory information, exchanging information, transferring cases, and conducting joint enforcement, and avoid unnecessary or duplicative inspections.

The national cyberspace administration, the telecommunications department under the State Council, and the public security authorities shall share, through information systems, information relating to internet information service approvals, internet connection filings, filings of multi-channel internet content distribution service organizations, filings of application distribution platforms, and the identities of registrants.

Where a violation is suspected to constitute a criminal offense, the relevant competent authority shall promptly transfer the case to the judicial authorities for criminal investigation in accordance with the law. Where criminal liability is not pursued or criminal punishment is exempted but administrative penalties remain appropriate, the judicial authorities shall promptly transfer the case to the relevant competent authority.

Article 67

Where, in accordance with these Measures, the telecommunications authority revokes, cancels, or withdraws a telecommunications business license, or cancels an approval number, it shall notify the relevant internet access service providers and domain name resolution service providers to cease providing services to the relevant entity.

Article 68

Any organization or individual that discovers conduct by an internet information service provider or internet access service provider in violation of these Measures may report or file a complaint with the relevant competent authority.

Upon receiving such reports or complaints, the cyberspace administration, telecommunications, public security, and other competent authorities shall promptly handle them in accordance with the law. Where the matter falls outside their jurisdiction, they shall promptly inform the complainant or transfer the matter to the competent authority.

Article 69

When conducting supervision and inspection of Platform Information Service Providers, the competent authorities shall focus on the following matters:

1. the handling and response to public complaints, reports, and user appeals;

2. the handling of unlawful internet accounts and information, the legal basis for such actions, and whether the platform itself has participated in any unlawful activities;

3. whether measures have been taken, in accordance with Article 25, paragraphs 1 and 2, to prevent and resist the production, reproduction, publication, dissemination, or transmission of undesirable information;

4. whether the provider has adopted measures required by the competent authorities to cooperate in the handling of unlawful or undesirable information;

5. the management of internet account credit records;

6. the formulation and amendment of platform rules and user agreements, including consultation procedures; and

7. any other matters requiring重点 supervision and inspection.

Where provincial-level or higher competent authorities, in the course of performing their supervisory duties, identify significant security risks or security incidents relating to internet information services, they may, in accordance with prescribed authority and procedures, summon the legal representative or principal person in charge of the relevant internet information service provider for an interview and require the provider to adopt measures to prevent the expansion of risks.

The internet information service provider shall take the required corrective measures and eliminate the identified risks.

Article 70

When conducting supervision and inspection in accordance with these Measures, the cyberspace administration, telecommunications, public security, and other competent authorities may, in accordance with the law, take the following measures:

1. question relevant parties and investigate matters relating to internet information services;

2. inspect and copy network logs and other materials relating to internet information services;

3. conduct on-site inspections and investigate suspected unlawful conduct; and

4. where electronic equipment, storage media, articles, facilities, or premises are connected with suspected unlawful conduct, adopt compulsory measures such as sealing or seizure in accordance with the law.

Article 71

State secrets, official secrets, trade secrets, personal privacy, and personal information obtained by competent authorities and their personnel in the course of supervising and administering internet information services shall be kept confidential. Such information may be used only for the purposes of supervision, administration, and law enforcement, and shall not be disclosed, altered, unlawfully destroyed, sold, or unlawfully provided to others.

Article 72

Providers of internet access services, internet information services, domain name registration services, domain name resolution services, and similar services shall provide the necessary support and assistance to the cyberspace administration and telecommunications authorities in the lawful performance of their supervisory responsibilities over internet information services.

Such providers shall also provide technical support and assistance to public security authorities and state security authorities in accordance with the law for activities relating to safeguarding national security and investigating criminal offenses.

Article 73

Where necessary to safeguard cybersecurity and social order or to respond to major cybersecurity incidents, and with the approval of the State Council, the national cyberspace administration, together with the telecommunications and public security departments under the State Council, may impose temporary control measures, including restrictions, on online applications or specific functions that possess the attributes of public opinion or the capacity for social mobilization within a specified period or geographic area.

Chapter V Legal Liability

Article 74

Where any person, in violation of Article 9 of these Measures, engages in telecommunications business without obtaining the required telecommunications business operating license, or provides services beyond the scope of the license, the telecommunications authority shall order correction, confiscate any unlawful gains, and impose a fine of three to five times the amount of the unlawful gains. Where there are no unlawful gains or the unlawful gains are less than RMB 50,000, a fine of RMB 100,000 to RMB 1,000,000 shall be imposed. In serious cases, suspension of business for rectification shall be ordered.

Where any person, in violation of Articles 8 or 17(2), provides internet information services without obtaining the required approval, or provides services beyond the approved scope, the telecommunications authority shall order correction. Where the provider refuses to make corrections, suspension of business for rectification shall be ordered.

Where any person fails to complete the required public security filing within the prescribed period in violation of Article 16(1), the public security authority shall issue a warning and order correction. Where the provider refuses to comply, suspension of network operations for rectification shall be ordered.

Where an internet information service provider violates Articles 13 or 14, the telecommunications authority shall order correction. Where the provider refuses to comply or the circumstances are serious, the authority may issue a warning, impose a fine of RMB 100,000 to RMB 1,000,000, suspend the relevant business, order suspension of operations for rectification, revoke the telecommunications business operating license, or cancel the approval number. The directly responsible persons and other directly liable personnel may be fined RMB 10,000 to RMB 100,000.

Where an internet information service provider engages in internet information services requiring prior approval under Article 11 without obtaining such approval, the cyberspace administration or other competent authorities shall, within the scope of their respective responsibilities, order the cessation of the relevant services, confiscate unlawful gains, and impose a fine of five to ten times the unlawful gains where such gains exceed RMB 10,000. Where unlawful gains are less than RMB 10,000 or there are no unlawful gains, a fine of RMB 10,000 to RMB 100,000 may be imposed. In serious cases, the telecommunications authority shall revoke the telecommunications business operating license or cancel the approval number.

Article 75

Where an internet information service provider obtains a license or approval through fraud, bribery, or other improper means in violation of Chapter II of these Measures, the original licensing or approving authority shall revoke the relevant license or cancel the approval number, confiscate unlawful gains, and may impose a fine of RMB 100,000 to RMB 1,000,000.

Article 76

Where an internet information service provider or internet livestreaming platform provider violates Articles 16(2), 17(1), 18, 34, 37(1), 41(1), 42, 43, 44, 57, 59, 62, or 72 of these Measures, the cyberspace administration, telecommunications authority, public security authority, or other competent authorities shall, within the scope of their respective responsibilities, issue a warning, order correction within a specified period, and confiscate unlawful gains.

Where the provider refuses to comply or the circumstances are serious, the competent authorities may impose a fine of RMB 50,000 to RMB 500,000, suspend the relevant business, order suspension of operations for rectification, close the website or application, revoke the relevant business license or business permit, and impose a fine of RMB 10,000 to RMB 100,000 on the directly responsible persons and other directly liable personnel.

Article 77

Where an internet information service provider violates Article 24, Article 25(1) or (2), Article 39, Article 54(1), Article 60, or Article 61(3), penalties shall be imposed in accordance with applicable laws and administrative regulations. Where no such provisions exist, the competent authorities, including the cyberspace administration, shall order correction, issue a warning, confiscate unlawful gains, and may impose a fine of RMB 50,000 to RMB 500,000, together with measures including suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of relevant licenses. The directly responsible persons and other directly liable personnel may be fined RMB 10,000 to RMB 100,000.

Where the provider refuses to make corrections or the circumstances are serious, the competent authorities shall confiscate unlawful gains, impose a fine of RMB 500,000 to RMB 2,000,000, and may order suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of relevant licenses. The directly responsible persons and other directly liable personnel shall be fined RMB 10,000 to RMB 100,000, and may additionally be prohibited from engaging in internet information security management activities for six months to one year, or have their relevant qualifications revoked.

Where the circumstances are especially serious, the competent authorities shall confiscate unlawful gains, impose a fine of RMB 2,000,000 to RMB 10,000,000, order suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of relevant licenses. The directly responsible persons and other directly liable personnel shall be fined RMB 100,000 to RMB 1,000,000, and may be prohibited from serving as directors, supervisors, senior executives, or persons responsible for internet information security in relevant enterprises for a specified period, and their internet information security qualifications may be revoked.

Article 78

Where providers of internet access services, domain name registration services, domain name resolution services, or similar services violate Articles 15, 34, or 65(2), the competent authorities, including the cyberspace administration, telecommunications authority, and public security authority, shall issue a warning, order correction within a specified period, and confiscate unlawful gains. Where the provider refuses to comply or the circumstances are serious, a fine of RMB 100,000 to RMB 500,000 may be imposed, together with suspension of business, suspension of operations for rectification, revocation of the relevant license, or revocation of the business permit. The directly responsible persons and other directly liable personnel may be fined RMB 10,000 to RMB 100,000.

Article 79

Where any person violates Articles 19, 20, 21, or 22, penalties shall be imposed in accordance with applicable laws and administrative regulations. Where no such provisions exist, the cyberspace administration, telecommunications authority, or public security authority shall issue a warning, order correction within a specified period, confiscate unlawful gains, and, where the provider refuses to comply or the circumstances are serious, impose a fine of RMB 50,000 to RMB 500,000, together with suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of relevant licenses. The directly responsible persons and other directly liable personnel may be fined RMB 10,000 to RMB 100,000.

Article 80

Where any person violates Article 31, Article 46, Article 47, Article 50, Article 51, Article 52, Article 53, Article 56, or Article 61(1) or (2) of these Measures, the relevant competent authorities shall order correction, issue a warning, confiscate unlawful gains, and may impose a fine of RMB 100,000 to RMB 1,000,000.

Where the provider refuses to comply or the circumstances are serious, a fine of RMB 1,000,000 to RMB 2,000,000 may be imposed, together with suspension of business, suspension of operations for rectification, or revocation of the relevant business license or business permit.

Article 81

Where an internet information service provider or internet access service provider violates Article 27, Article 28, Article 29(1), Article 40(2), Article 63, or Article 64 of these Measures, the relevant competent authorities shall order correction, issue a warning or a public criticism, and may impose a fine of RMB 50,000 to RMB 500,000. They may also order suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of the relevant business license or permit.

Where the provider refuses to comply or the circumstances are serious, a fine of RMB 500,000 to RMB 2,000,000 may be imposed, together with suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of the relevant business license or permit. The directly responsible persons and other directly liable personnel may be fined RMB 50,000 to RMB 200,000.

Where the violations referred to in the preceding paragraph cause particularly serious impacts or consequences, the competent authorities may impose a fine of RMB 2,000,000 to RMB 10,000,000, order suspension of business, suspension of operations for rectification, closure of websites or applications, or revocation of the relevant business license or permit. The directly responsible persons and other directly liable personnel may be fined RMB 200,000 to RMB 1,000,000.

Article 82

Where the cyberspace administration, telecommunications, public security, or other competent authorities, or their personnel, use information obtained in the course of supervising internet information services for purposes other than those permitted under Article 71 of these Measures, the responsible leadership personnel and directly responsible personnel shall be subject to disciplinary sanctions in accordance with the law.

Where personnel of the cyberspace administration, telecommunications, or public security authorities neglect their duties, abuse their powers, engage in malpractice for personal gain, or solicit or accept property by taking advantage of their official positions, disciplinary sanctions shall be imposed in accordance with the law.

Article 83

Where an operator of an internet public account violates Article 24 of these Measures, penalties shall be imposed in accordance with applicable laws and administrative regulations. Where no such provisions exist, the cyberspace administration or other competent authorities shall, within the scope of their respective responsibilities, issue a warning, order correction within a specified period, require the removal of subscribers where appropriate, restrict account functions, suspend or restrict monetization privileges, suspend relevant services, close the account, or prohibit re-registration, and may impose a fine of RMB 10,000 to RMB 50,000.

Where an operator of an internet public account with significant influence violates Article 45 of these Measures, the competent authorities shall issue a warning, order correction within a specified period, require the removal of subscribers where appropriate, restrict account functions, suspend or restrict monetization privileges, suspend relevant services, close the account, or prohibit re-registration, and may impose a fine of RMB 50,000 to RMB 500,000.

Article 84

Where any person violates Article 26, Article 30, Article 32, Article 48, or Article 49 of these Measures, penalties shall be imposed in accordance with applicable laws and administrative regulations. Where no such provisions exist, the cyberspace administration, telecommunications, public security, or other competent authorities shall, within the scope of their respective responsibilities, order correction within a specified period and may impose a fine of RMB 100,000 to RMB 1,000,000.

Where the provider refuses to comply or the circumstances are serious, a fine of RMB 1,000,000 to RMB 5,000,000 may be imposed, together with suspension of business or suspension of operations for rectification.

Article 85

Where an internet information service provider violates Article 33(1) of these Measures, the telecommunications authority, cyberspace administration, or other competent authorities shall order correction within a specified period and may impose a fine of RMB 100,000 to RMB 1,000,000.

Where the provider refuses to comply or the circumstances are serious, a fine of RMB 1,000,000 to RMB 2,000,000 may be imposed, together with suspension of business or suspension of operations for rectification.

Article 86

Where any person violates other provisions of these Measures, legal liability shall be pursued in accordance with the Cybersecurity Law of the People’s Republic of China, the Data Security Law of the People’s Republic of China, the Personal Information Protection Law of the People’s Republic of China, the Anti-Unfair Competition Law of the People’s Republic of China, the Law of the People’s Republic of China on the Protection of Consumer Rights and Interests, the Regulations on the Online Protection of Minors, and other applicable laws and administrative regulations.

Article 87

Where a violation of these Measures causes damage to another person, civil liability shall be borne in accordance with the law. Where the conduct constitutes a violation of public security administration, public security administrative penalties shall be imposed in accordance with the law. Where the conduct constitutes a criminal offense, criminal liability shall be pursued in accordance with the law.

Article 88

Where conduct in violation of these Measures is included, in accordance with laws, administrative regulations, or relevant national provisions, in the List of Seriously Dishonest Entities in the Internet Sector, the relevant competent authorities may impose disciplinary measures for dishonesty, including restricting or prohibiting the registration of new user accounts, restricting account functions, or prohibiting the entity from engaging in internet information services for a specified period or within a specified scope.

Article 89

Where a violation of these Measures falls within the circumstances warranting mitigated punishment, reduced punishment, or exemption from administrative penalties under the Administrative Penalty Law of the People’s Republic of China, the relevant provisions of that Law shall apply.

Chapter VI Supplementary Provisions

Article 90

For the purposes of these Measures, the following terms shall have the meanings set out below:

1. Internet information services refer to activities that provide information to the public through internet websites, applications, AI agents, microblogs, instant messaging tools, search engines, internet livestreaming, online audiovisual services, image and text sharing, knowledge learning and reasoning, task assistance and execution, internet public accounts, and other forms.

2. Internet access services refer to services that provide network access for internet information service providers, including internet data center (IDC) services, content delivery network (CDN) services, internet access services, and other business forms, such as proxy services, server hosting, and space leasing.

3. Internet accounts refer to accounts registered and used by users of internet information services.

4. Internet public accounts refer to internet accounts used by internet information service users to produce and publish text, images, audio, video, or other information for the general public, excluding accounts that provide only communication functions.

5. Internet public accounts with significant influence refer to internet public accounts that meet one or more of the following criteria: the number of followers, views, reposts, or comments reaches a specified scale; revenue generated through the account during the previous year reaches a specified scale; or the account has significant influence within a particular sector or community.

6. Operators of internet public accounts refer to organizations or individuals that register and operate internet public accounts to produce and publish information.

7. Multi-channel internet content distribution services refer to services that provide planning, production, publication, distribution, marketing, promotion, agency, or similar services for information published through internet public accounts.

8. Internet livestreaming platform service providers refer to Platform Information Service Providers that provide internet livestreaming services.

9. Application distribution platform service providers refer to Platform Information Service Providers that provide application distribution services.

10. Large internet platforms refer to internet platforms with more than 50 million registered users or 10 million monthly active users, whose business operations are complex and whose internet information services have a significant impact on economic activities, online order, or users’ rights and interests.

Article 91

Where a foreign-invested enterprise engages in internet information services, it shall also comply with the requirements of laws, administrative regulations, and relevant national provisions governing foreign investment.

Where laws, administrative regulations, or relevant national provisions contain special rules governing foreign-invested enterprises engaged in e-commerce retail, e-commerce supply chain services, or other e-commerce activities, those provisions shall apply.

Article 92

Organizations or individuals outside the territory of the People’s Republic of China that provide internet information services to users within China, or make use of internet information services within China, shall comply with laws, administrative regulations, and relevant national provisions.

Where an international treaty concluded or acceded to by the People’s Republic of China contains provisions different from these Measures, the treaty shall prevail, except for provisions to which China has entered reservations.

Article 93

Where internet information services involve network operational security, data security, personal information protection, anti-unfair competition, consumer protection, news dissemination, or media administration, the relevant laws and administrative regulations shall also apply.

Where internet information services involve State secrets, the relevant laws and administrative regulations on the protection of State secrets shall also apply.

Article 94

These Measures shall enter into force on [date to be determined].

Explanatory Notes on the Draft Revision of the Measures for the Administration of Internet Information Services

The CPC Central Committee and the State Council attach great importance to the administration of internet information services and have made important arrangements for revising the Measures for the Administration of Internet Information Services (the “Measures”).

In accordance with the State Council’s legislative work plan, the Cyberspace Administration of China (”CAC”), together with the Ministry of Industry and Information Technology (”MIIT”), the Ministry of Public Security (”MPS”), and other relevant authorities, conducted extensive legislative research, carefully reviewed practical experience, and broadly solicited opinions from all stakeholders. On this basis, the authorities prepared the Draft Revision of the Measures for the Administration of Internet Information Services (the “Draft”) and hereby release it for public comment.

The relevant matters are explained as follows.

I. Necessity of the Revision

As a foundational regulation governing internet information services, the Measures have played an important role since their implementation in 2000 in regulating the online information environment and promoting the sustained and healthy development of China’s internet industry.

However, with the rapid development of information technology, profound changes have taken place in the governance structure, regulatory mechanisms, regulatory tools, and regulatory targets of internet information services. Accordingly, the existing Measures require further revision and improvement.

First, the revision implements the decisions and arrangements of the CPC Central Committee.

General Secretary Xi Jinping has emphasized the need to improve the comprehensive governance framework for cyberspace under the centralized and unified leadership of the Party, and to strengthen guidance over online platforms, self-media accounts, and multi-channel network (MCN) organizations so that they fulfill their social responsibilities.

The 15th Five-Year Plan for National Economic and Social Development also calls for strengthening online content governance and improving laws and regulations governing internet content management and platform governance.

Second, the revision responds to new challenges in internet information governance.

In recent years, a range of new issues have emerged in the internet information services sector. Some self-media accounts and online public relations firms have sought improper commercial gains by providing paid influence services or generating artificial traffic. Some online platforms have abused platform rules or engaged in algorithmic discrimination, thereby infringing users’ lawful rights and interests.

Meanwhile, the rapid adoption of technologies such as generative artificial intelligence and deep synthesis has significantly increased the complexity of internet governance.

There is therefore an urgent need to enrich the legal basis for internet governance, address prominent regulatory challenges, and establish a stronger institutional foundation for internet information services under the new technological environment.

Third, the revision aligns the Measures with recently enacted legislation.

In recent years, China has enacted or amended a number of important laws, including the Cybersecurity Law, the Law on the Protection of Heroes and Martyrs, the Personal Information Protection Law, and the Law on the Protection of Minors. In addition, the CAC and other authorities have issued numerous departmental rules and normative documents covering online ecosystem governance, internet account administration, cyberbullying, and the development and security of AI-enabled information services.

The revision therefore seeks to better align the Measures with this broader legislative framework while elevating mature regulatory practices into a higher-level administrative regulation.

II. Revision Process

The CAC, together with MIIT, MPS, and other relevant authorities, carried out extensive legislative research and systematically reviewed China’s experience in regulating internet information services.

Based on this work, the drafting team focused on key challenges in internet governance, proposed new regulatory mechanisms where necessary, aligned the Draft with relevant legislation, and organized consultations with experts and scholars.

The Draft was also circulated for comments among central government agencies, provincial cyberspace administrations, and the Cyberspace Administration of the Xinjiang Production and Construction Corps.

Following further revisions and the completion of the required legislative assessments, the current Draft was finalized for public consultation.

III. Overall Approach

The revision follows four guiding principles.

First, it upholds the leadership of the Communist Party of China.

The Draft is guided by Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era—particularly Xi Jinping Thought on the Rule of Law and Xi Jinping Thought on Culture—and implements the decisions of the 20th CPC National Congress and subsequent plenary sessions, as well as General Secretary Xi Jinping’s important instructions on internet information service governance.

Second, it follows a problem-oriented approach.

The Draft is grounded in China’s practical experience in internet governance, responds to emerging regulatory challenges, improves institutional design, aligns with existing legislation, and strengthens the systematic, targeted, and practical nature of the regulatory framework.

Third, it emphasizes collaborative governance.

The Draft further clarifies the respective responsibilities of government authorities, industry associations, internet information service providers, operators of internet public accounts, industry practitioners, and other relevant participants.

Fourth, it balances continuity and innovation.

Regulatory mechanisms that have proven effective under the existing Measures are retained and improved, while provisions that no longer reflect technological or market realities have been revised, supplemented, or removed in light of developments in the information technology sector.

IV. Major Revisions

The Draft consists of six chapters and ninety-four articles, representing a comprehensive revision of the current Measures.

The principal revisions include the following.

First, the Draft clarifies the fundamental principles and regulatory framework governing internet information services.

It expressly provides that internet information services and their supervision shall uphold the leadership of the Communist Party of China and be guided by the core socialist values. It also clarifies the respective responsibilities of the CAC, the telecommunications authorities under the State Council, the Ministry of Public Security, and other competent authorities, while specifying that local supervisory responsibilities shall be determined in accordance with relevant national provisions.

Second, the Draft strengthens the regulation of internet information service providers.

It further improves market entry requirements, standardizes the use of network resources, and enhances requirements relating to industry personnel. It requires providers to obtain approval from the telecommunications authorities before providing internet information services, to use network resources that meet regulatory requirements, and requires personnel engaged in internet news information services to obtain relevant qualifications and complete prescribed training and assessments.

Third, the Draft strengthens regulation of internet information service activities.

It further enhances user account administration, improves rules governing online information content, and introduces new regulatory requirements for specific circumstances.

Among other things, it requires dynamic verification of registered accounts, prohibits the production, publication, or dissemination of unlawful information, requires providers to prevent and resist undesirable information, prohibits conduct that disrupts the order of internet information services, and prohibits providing support or assistance for unlawful online activities.

Fourth, the Draft strengthens the responsibilities of online platforms.

It introduces a dedicated chapter on Platform Information Services, requiring operators of internet public accounts with significant influence not to produce or disseminate unlawful or undesirable information.

It also requires platforms to strengthen the identification and monitoring of cyberbullying, adopt timely intervention measures where cyberbullying risks arise, provide users with evidence preservation tools, and regulate multi-channel internet content distribution organizations and the public accounts operating under them.

Fifth, the Draft promotes the secure development of intelligent information services.

A dedicated chapter on Intelligent Information Services establishes a comprehensive regulatory framework for AI-enabled internet services.

The Draft strengthens requirements relating to the labeling of AI-generated content, regulates the use of recommendation algorithms, prohibits forcing users to use AI services, prohibits the use of algorithms to manipulate online public opinion or evade regulatory oversight, strengthens protection of workers affected by algorithmic management, and enhances mechanisms for responding to significant AI-related security risks.

Finally, the Draft improves the legal liability regime and expands the regulatory toolbox available to enforcement authorities.

Drawing upon recent experience in internet governance, it establishes a List of Seriously Dishonest Entities in the Internet Sector, introduces additional enforcement measures such as restricting account functions and prohibiting the registration of new user accounts, aligns with the Cybersecurity Law, the Law on the Protection of Minors, the Administrative Penalty Law, and other relevant legislation, and further strengthens the legal liability framework.

LongCat-2.0, released today by Meituan, is a 1.6 trillion-parameter MoE model with about 48 billion parameters activated per token, trained on more than 30 trillion tokens and built for agentic coding. It supports a 1 million-token context window, and Meituan says the full pretraining run and large-scale deployment were completed on a 50,000-card cluster of domestic Chinese chips, then quietly tested in the open as Owl Alpha on OpenRouter before the company revealed what it was.

That is already enough to make it notable. Most discussion around export controls has focused on whether Chinese labs can continue to train frontier-scale models without access to the latest Nvidia GPUs. Meituan's announcement does not settle that argument, but it does narrow it. A company better known outside China for food delivery and local services is now saying, in plain terms, that it doesn’t need Nvidia GPUs to run a trillion-parameter-class model both in training and in inference.

Meituan's own announcement says LongCat-2.0 was trained from scratch, used more than 30 trillion tokens of pretraining data, and ran on a domestic cluster where the team had to solve communication faults, memory pressure, numerical stability, deterministic operators, and distributed recovery at very large scale. The LongCat blog adds another number: more than 35 trillion tokens consumed during the full training run and deployment validation, again on domestic chips, with no rollback and no unrecoverable loss spike.

They imply that the harder problem was not merely getting a single large run to start, but keeping a large system stable enough to finish. Meituan also says the architecture was shaped around agentic coding: sparse attention for long context, dynamic token-level compute allocation, and an MoE layout designed for coding, reasoning, and interaction tasks rather than generic chatbot performance.

This is an important distinction. The company is not presenting LongCat-2.0 as a universal answer to every benchmark. It is presenting it as a model meant to work in code understanding, repository-scale modification, tool use, and multi-step execution. That is a narrower claim, but a more commercially legible one.

On June 29, the official Meituan LongCat account on X wrote: "Owl Alpha on OpenRouter — that's us." The post added that Owl Alpha had reached the global top three by daily volume on OpenRouter, while ranking first on Hermes Agent, second on Claude Code, and third on OpenClaw by monthly volume. It also said the stealth phase was ending and Owl Alpha would retire soon.

That single sentence explains why the LongCat launch felt oddly familiar to developers who had been spending time around coding agents. They had already been using the model. They just had not been told what it was.

Rather than introducing a branded model and then waiting for adoption, Meituan appears to have done the reverse. It let the model circulate in a neutral developer environment under a different name, watched whether people actually chose it for real work, and only later attached the company brand. By the time the identity was disclosed, the model already had usage, rankings, and a reputation of its own.

It would be wrong, though, to say that Meituan had been secretly absent from the open-source community until now. The stronger version of the claim is more interesting: the team had been visible for some time, but its flagship coding model was not.

The Hugging Face organization meituan-longcat explicitly identifies itself as Meituan LongCat. Its page shows a continuing release history across models, datasets, benchmarks, spaces, and technical reports. LongCat-2.0 appeared there as a new model update, but it sits alongside earlier projects such as LongCat-Next, LongCat-Flash-Prover, LongCat-Video, LongCat-AudioDiT, WBench, LARYBench, MineExplorer, and LoHoSearch.

The meituan-longcat organization has 35 public repositories on GitHub as well, including LongCat-2.0, LongCat-Flash-Chat, LongCat-Next, LongCat-Video, LongCat-Image, and various supporting components. The organization page also says it has no public members, which is a small but telling detail. The code and assets were public; the people behind them were kept in the background.

The arXiv record shows LongCat-Flash, LongCat-Flash-Thinking, LongCat-Flash-Omni, LongCat-Next, and LongCat-Video technical reports across text, reasoning, multimodal interaction, and generative media. There were public model cards and repositories. There were datasets and benchmarks. There was an API platform with a visible update history, including the release of LongCat-Flash-Chat and later the rollout of LongCat-2.0-Preview.

Seen this way, Owl Alpha was not a detour from the LongCat strategy. It was a natural extension of it. The public repositories built continuity. The anonymous OpenRouter deployment built proof of use. The June 2026 reveal tied the two together.

LongCat-2.0 does not invalidate the logic behind U.S. export controls. Restrictions still raise cost, slow access, complicate scaling, and force Chinese firms into harder engineering trade-offs. But the Meituan case does put pressure on one of the simpler assumptions behind the policy: that denying the newest Nvidia stack would prevent Chinese actors from training and serving frontier-adjacent systems at very large scale.

What Meituan is really showing is that the binding constraint has shifted. The old question was whether domestic Chinese hardware could, in principle, replace imported accelerators at the high end. The newer question is whether domestic hardware, plus enough systems engineering, can support serious models for real users. Those are not the same question. LongCat-2.0 is interesting because it tries to answer the second one.

For all the specificity around cluster size and training scale, one key piece is still missing: Meituan has not publicly named the chip supplier or the chip model.

The official language stops at "domestic chips" and "domestic compute." Mainstream coverage in outlets such as Xinhua, People's Posts and Telecommunications News, Sina Tech, and IT Home repeats that phrasing without identifying a vendor. The LongCat blog is equally careful. It says domestic chips, not Ascend, not Muxi, not Suiyuan, not Biren, not Cambricon.

That silence has produced the usual market guessing. Huawei Ascend is the most common candidate in public discussion, largely because it is the ecosystem most people associate with large-scale domestic training. But that remains a guess. There is no official confirmation of supplier or SKU in the material now available. The most accurate sentence one can write at this point is a narrow one: Meituan has publicly claimed full training and inference on domestic Chinese chips, but it has not publicly disclosed which chips those were.

That may be the real significance of LongCat-2.0. A company most outsiders still file under food delivery has shown that, in 2026, frontier-scale AI work can emerge from places that American policy and foreign investors were not watching closely enough.

Interestingly, over the past few days, Meituan's management has made a series of unusually coordinated public efforts to reassure investors. With the company's share price down more than 30% year-to-date and its market capitalization falling below HK$400 billion, CEO Wang Xing acknowledged at the annual shareholders' meeting that the stock's performance over the past few years had been "unsatisfactory" and said he bore "significant responsibility" for it. He also stressed that he has never sold a single share of Meituan since the company was founded, adding that the only change in his holdings, in 2021, resulted from a donation of shares to a charitable foundation rather than any cash-out.

At the same time, Meituan’s CFO Chen Shaohui argued that Meituan is "significantly undervalued" by the market, saying the company plans to resume share buybacks and will gradually monetize parts of its investment portfolio, returning the proceeds to shareholders.

If Meituan wants LongCat-2.0 to matter beyond one news cycle, three things will need to happen. First, the community will need enough access to the weights, reports, and tooling to test whether the architecture and performance claims travel well outside Meituan's own environment. Second, the OpenRouter episode will need to translate into durable developer adoption under the LongCat name, not just curiosity about the unmasking of Owl Alpha. Third, any future disclosure around the chip stack will need to come from official or at least clearly attributable reporting, because right now the supplier question remains open.

MOFCOM also clarified that exports of dual-use items to entities on the Watch List will be subject to enhanced end-user and end-use reviews.

Newly Added Entities to the Export Control List:

Newly Added Entities to the Watch List:

Implications for Entities Added to the Watch List

Entities placed on the Watch List will no longer be eligible for General Licenses or the Registration Filing Pass. Applications for individual export licenses must be accompanied by a written non-military end-use commitment, significantly increasing compliance costs. In addition, license reviews are not subject to the statutory review timeline under Article 17 of the Regulations, meaning approvals can, in practice, be delayed indefinitely.

As a result, entities involved in the nuclear sector (e.g., Mitsubishi Nuclear Fuel and Japan Nuclear Fuel Limited), the defense sector (e.g., Howa Machinery, Hosoya Pyro-Engineering, Fujikura Parachute, and YDK), and the drone sector (e.g., ACSL and Terra Drone) are likely to face persistent uncertainty in contract execution, deliveries, and overseas maintenance and support.

However, compared with entities on the Export Control List, those on the Watch List retain access to a potential export channel because they may still apply for individual export licenses. They may also apply for removal from the Watch List after fulfilling the required verification and compliance obligations. Moreover, the routine civilian operations of non-defense subsidiaries within the same corporate group are, in principle, not directly restricted by the listing.

Notably, several of the newly listed entities have existing operations or commercial ties in China.

For the OKI group, five affiliated entities were listed at the same time. OKI also maintains manufacturing, service, and software delivery sites in Shenzhen, Changzhou, and Dalian. This round of listings could therefore create direct compliance pressure on its internal flow of goods and technology under a “Japan headquarters receives the order → China-based production/R&D supports delivery” model.

For the Komatsu group, sales in China declined by 5.5% year on year, but its subsidiary Komatsu NTC only converted its China base into a wholly owned operation in 2022, suggesting that China remains an important production node for the Asian market. The listing will raise the compliance threshold for any China-origin components needed for semiconductor equipment that are supplied back to the Japanese parent.

For the drone sector, including ACSL and Terra Drone, there was already a precedent in early 2026 where imports were stalled because China-made motors were treated as drone-related components. Following their addition to the Watch List, supply-chain substitution pressure for Japan’s domestically produced industrial drones is likely to move further upstream.

Where China’s leverage lies: the transmission mechanism through rare earths and dual-use items:

Timeline of China’s Sanctions Measures Since Takaichi’s Remarks

On November 7, 2025, Japanese Prime Minister Sanae Takaichi stated during a Diet session that, if mainland China were to use force against Taiwan, it could constitute a “survival-threatening situation” under Japanese law, implying that the Self-Defense Forces could exercise the right of collective self-defense and intervene.

Six days later, on the evening of November 13, Chinese Vice Foreign Minister Sun Weidong, acting on instructions, summoned Japanese Ambassador to China Kenji Kanasugi and lodged solemn representations over Japan’s erroneous remarks and actions concerning China. This marked the point at which the issue formally entered a phase of bilateral diplomatic confrontation.

Following the escalation of tensions, China first used stricter export controls on dual-use items to Japan as a preparatory measure. On January 6, 2026, MOFCOM issued Announcement No. 1 of 2026, prohibiting the export of all dual-use items to Japanese military end-users, for military end-uses, and for “any other end-users or end-uses that contribute to enhancing Japan’s military capabilities.” Third-party reports said this round covered approximately 1,030 controlled items.

On February 24, 2026, pursuant to the Export Control Law and the Regulations on Export Control of Dual-Use Items, MOFCOM issued Announcements No. 11 and No. 12, respectively adding 20 entities, including Mitsubishi Shipbuilding, to the Export Control List, and another 20 entities, including Subaru, to the Watch List. MOFCOM stated that the purpose was to safeguard national security and interests and fulfill international obligations such as non-proliferation.

The June 29 listings represent China’s latest action following the two earlier rounds of sanctions.

The 80 Japanese entities sanctioned to date fall almost entirely into three clusters:

1. Defense and aerospace

   
   Including Mitsubishi Heavy Industries affiliates, Kawasaki Heavy Industries affiliates, IHI affiliates, JAXA, the National Defense Academy, the National Institute for Defense Studies, Mitsubishi Electric defense/space units, Japan Aircraft, and related defense contractors.

2. Shipbuilding, engines, heavy machinery and semiconductor equipment

   
   Including Mitsubishi Shipbuilding, Japan Marine United, JMU Defense Systems, Mitsui E&S, Komatsu Industries, Komatsu NTC, and related industrial machinery entities.

3. Dual-use components, drones, nuclear and advanced materials

   
   Including Subaru, TDK, Mitsubishi Materials, Nitto Denko, NOF, Terra Drone, ACSL, Mitsubishi Nuclear Fuel, Japan Nuclear Fuel, Howa Machinery, Hosoya Pyro-Engineering, and Fujikura Parachute.

Another noteworthy development occurred on the night of June 22–23, when numerous Chinese self-media accounts simultaneously circulated claims that Japan had halted exports of photoresists to China.

A large number of Chinese-language articles used nearly identical wording, specifically naming the four leading Japanese photoresist suppliers—Tokyo Ohka Kogyo (TOK), JSR, Shin-Etsu Chemical, and Fujifilm Electronic Materials—and claiming that they had already implemented the following measures:

• stopped accepting all new ArF and EUV photoresist orders;

• significantly reduced new KrF photoresist orders;

• withdrawn all resident field engineers from customer sites; and

• extended delivery times for existing photoresist contracts from 1–2 months to 6–8 months.

However, neither the Japanese government nor any of the four companies issued an official announcement to that effect. Moreover, no reports appeared in major media outlets, including Nikkei, Kyodo News, Bloomberg, or Reuters. Within a few hours, most of the articles were either taken down or their links became inaccessible.

To understand its significance, it helps to look at the broader timeline.

On March 31, 2026, Premier Li Qiang signed State Council Decree No. 834, the Regulations on Industrial and Supply Chain Security. Consisting of just 18 articles, it became China’s first dedicated administrative regulation governing industrial and supply chain security.

A week later, on April 7, Ministry of Justice officials described the regulation as filling a long-standing legislative gap and indicated that a dedicated “industrial and supply chain security investigation mechanism” would be one of its key implementing tools.

On June 22, MOFCOM formally issued the implementing measures.

The relationship between the two documents is straightforward. Decree 834 serves as the framework legislation, granting the government authority to investigate actions that threaten China’s industrial and supply chain security. The newly issued Measures provide the operating manual, explaining how investigations will be initiated, conducted, and ultimately enforced. In practical terms, this marks the first time China has established a standing enforcement mechanism specifically designed to respond to foreign actions deemed harmful to China’s industrial and supply chains.

Compared with the broad principles contained in Decree 834, the Measures provide several important layers of detail.

First, they identify the lead authority. While the State Council regulation referred only to “relevant departments of the State Council,” Article 2 of the Measures explicitly designates MOFCOM as the responsible agency.

Second, they define both the criteria for assessing harm and the channels for initiating investigations. Decree 834 granted investigative authority but did not explain how harm would be evaluated. Article 4 of the Measures fills that gap by identifying four categories of factors, including impacts on the security of critical materials, technologies, capital, assets, data, information, and personnel; impacts on the flow of logistics, trade, people, capital, data, and information; impacts on international competitiveness and development potential; and other relevant factors. What was previously a broad concept of “harm to industrial and supply chain security” is thus transformed into a more operational framework for regulators, companies, and legal advisers.

Articles 5 and 6 also establish formal reporting channels. Chinese companies and organizations may submit evidence and reports directly to MOFCOM or through provincial commerce authorities. In effect, this creates China’s first formal industrial-relief mechanism specifically aimed at foreign measures that disrupt supply chains, somewhat analogous to how companies petition the U.S. International Trade Commission in trade remedy cases.

Third, and most importantly, the Measures establish a detailed investigative process. While Decree 834 merely referenced tools such as interviews and document reviews, Articles 7 through 16 create a complete procedural framework. Article 7 covers case initiation and public notice requirements. Article 8 expands investigative methods to include interviews, document inspection and copying, public solicitation of information, questionnaires, sampling, technical assessments, hearings, and on-site investigations. Article 9 authorizes expert advisory groups. Article 11 guarantees rights of explanation and defense. Article 12 contains a particularly consequential provision: if a target refuses to cooperate, fails to provide information, misses deadlines, or otherwise obstructs an investigation, MOFCOM may make determinations based on available facts.

Among these provisions, Article 10 stands out. It states that MOFCOM may conduct on-site investigations in foreign countries or regions when necessary, provided the relevant jurisdiction does not object. In principle, this extends Chinese investigative authority overseas, while formally preserving respect for sovereignty and territorial jurisdiction. In practice, few governments are likely to permit such investigations, but the provision itself signals Beijing’s growing willingness to project enforcement authority beyond its borders.

Fourth, the Measures establish mechanisms for consultation, suspension, and settlement. Investigations are not designed to be purely punitive. Articles 13 through 16 allow MOFCOM to engage in consultations, suspend investigations, terminate investigations if harmful measures are withdrawn or solutions are reached, and issue final determinations. This creates a flexible framework more akin to a policy instrument than a purely legal proceeding.

Fifth, the Measures establish a layered enforcement structure and domestic compliance mechanism. Articles 17 through 20 elaborate on the response authorities originally provided under Decree 834.

Against foreign governments, regions, or international organizations, China may impose import and export restrictions, special fees, inclusion on countermeasure lists, and other measures.

Against foreign companies and individuals, China may restrict imports, exports, investment, business cooperation, entry, stay, and residence. These restrictions may also extend to entities they control or participate in operating.

Against domestic entities that fail to comply, authorities may first order corrective action and then impose restrictions involving government procurement, public tenders, imports and exports, cross-border data transfers, and other activities.

Article 20 further authorizes MOFCOM to adjust measures dynamically over time.

Several provisions deserve particular attention.

The first is Article 17(2), which authorizes the collection of “special fees.” This concept has few direct precedents in Chinese legislation. The closest example was the special port fees imposed in 2025 on certain U.S.-related vessels in response to U.S. Section 301 measures. The provision effectively creates a mechanism for imposing targeted fees without relying on China’s tariff law. In the future, it could potentially serve as the legal basis for charging security-related fees on specific foreign services, cross-border data transfers, or other activities. Notably, the provision is not limited to goods or country-of-origin concepts. It applies to countries, regions, or organizations that implement or assist actions deemed harmful to China’s supply chain security, and could theoretically cover shipping, ports, certification, licensing, services trade, financial settlement, digital platforms, and other sectors.

The second is Article 18, which functions as a Chinese version of a “50 percent rule.” Measures may apply not only to targeted foreign entities themselves, but also to organizations they actually control or participate in establishing or operating. Unlike U.S. sanctions practice, however, the provision does not rely on specific ownership thresholds. Concepts such as actual control, participation in establishment, and participation in operations leave considerable room for interpretation and could potentially encompass joint ventures, contractual arrangements, and business partners.

The third is Article 19, which effectively creates a secondary compliance obligation. Domestic entities that continue dealing with sanctioned foreign parties may themselves face restrictions involving government procurement, cross-border data transfers, imports and exports, and other activities. The logic resembles U.S. secondary sanctions, although directed toward compliance obligations of domestic Chinese entities. Once measures are imposed under Articles 17 or 18, Chinese customers, suppliers, and business partners may be required to disengage or face consequences themselves.

For foreign companies, the implications are potentially significant. Previously, China’s Unreliable Entity List was primarily triggered by disruptions to normal transactions with Chinese firms. The new framework is broader. Foreign entities may now face investigation if they are deemed to have implemented discriminatory restrictions or measures inconsistent with international law and basic norms of international relations, even when acting pursuant to their own domestic legal obligations. This narrows the scope for the traditional defense that a company was merely complying with home-country law.

Taken together, Articles 17 and 18 create a toolkit that could simultaneously restrict imports, exports, investment, transactions, personnel entry, and work authorization. The scope of potential remedies exceeds anything previously available under China’s existing enforcement mechanisms.

Despite the significance of the Measures, several key questions remain unanswered.

First, will the “critical sectors list” envisioned under Article 11 of Decree 834 be published, and if so, which industries will it cover? This may ultimately prove even more consequential than the investigation mechanism itself.

Second, who will become the target of the first investigation? Historically, the first case often sets the tone for how an entire enforcement system is used.

Third, how will the “special fees” authority be implemented in practice? The provision could become China’s institutionalized equivalent of certain Western tariff and fee-based countermeasures, but substantial implementation details remain undefined.

Fourth, how will the new framework interact with the Unreliable Entity List? The triggering conditions overlap significantly, yet the Measures do not clarify the relationship between the two systems.

Finally, how will China defend the framework under WTO rules? Beijing has long criticized U.S. Section 301 investigations and other unilateral measures. Now that China has created a structurally similar instrument of its own, questions regarding WTO compatibility and legal justification are likely to become increasingly important.

Measures for Industrial and Supply Chain Security Investigations

Article 1

These Measures are formulated in accordance with the National Security Law of the People’s Republic of China, the Foreign Relations Law of the People’s Republic of China, the Anti-Foreign Sanctions Law of the People’s Republic of China, the Foreign Trade Law of the People’s Republic of China, the Regulations of the State Council on Industrial and Supply Chain Security, and other relevant provisions, for the purpose of conducting industrial and supply chain security investigations, preventing industrial and supply chain security risks, and enhancing the resilience and security of industrial and supply chains.

Article 2

The Ministry of Commerce (MOFCOM) shall be responsible for industrial and supply chain security investigations and shall carry out investigative activities accordingly.

Article 3

MOFCOM may conduct industrial and supply chain security investigations into the following measures or acts:

Foreign states, regions, and international organizations that violate international law and the basic norms governing international relations by adopting discriminatory prohibitions, restrictions, or similar measures against China in the field of industrial and supply chains, or by implementing or assisting in acts that harm the security of China’s industrial and supply chains;

Foreign organizations or individuals that violate normal market transaction principles by interrupting normal transactions with Chinese citizens or organizations, imposing discriminatory measures on Chinese citizens or organizations, or engaging in other acts that cause substantial harm or pose a substantial threat to the security of China’s industrial and supply chains.

Article 4

In assessing harm to industrial and supply chain security, MOFCOM may consider the following factors:

The impact of the relevant measures or acts on the security of domestic and overseas industrial and supply chain factors, including important materials, technologies, capital, assets, data, information, personnel, enterprises, and projects;

The impact of the relevant measures or acts on the smooth flow of domestic and overseas industrial and supply chain factors, including logistics flows, commercial flows, personnel flows, capital flows, data flows, and information flows;

The impact of the relevant measures or acts on the international competitiveness and development potential of China’s industrial and supply chains;

Other circumstances affecting industrial and supply chain security.

Article 5

Where domestic legal persons, organizations, or other entities believe that measures or acts undertaken by foreign states, regions, organizations, individuals, or international organizations have harmed or may harm the security of China’s industrial and supply chains, and that an industrial and supply chain security investigation should be initiated, they may submit written materials, including evidence and reports, to MOFCOM.

Article 6

Domestic legal persons, organizations, or other entities may submit relevant evidence, reports, and other written materials to MOFCOM through the commerce authorities of provinces, autonomous regions, municipalities directly under the Central Government, cities under separate state planning, and the Xinjiang Production and Construction Corps.

Article 7

MOFCOM may, based on actual circumstances, decide to initiate a case and publish a notice of investigation.

Article 8

MOFCOM may conduct investigations through means including interviews with investigated parties and other relevant persons, inspection or copying of relevant documents and materials, public solicitation of information, questionnaires, sampling surveys, technical assessments, hearings, and on-site investigations. Relevant parties shall cooperate accordingly.

MOFCOM may also proactively collect relevant information during the course of an investigation.

Article 9

Where deemed necessary, MOFCOM may establish an expert advisory panel to provide opinions on technical and legal issues arising during an investigation.

Members of the expert advisory panel shall properly safeguard confidential materials and information obtained during the investigation and shall bear corresponding confidentiality obligations.

Article 10

Where deemed necessary, MOFCOM may conduct on-site investigations in relevant foreign countries or regions, except where the relevant country or region raises objections.

Article 11

Relevant parties, including the investigated party, may make statements and present defenses to MOFCOM during the course of an investigation.

MOFCOM may require relevant parties to submit or supplement written materials. Relevant parties may also voluntarily submit written materials.

Article 12

Where an investigated party fails to cooperate with an investigation, fails to truthfully provide information or relevant materials within a reasonable period, or otherwise seriously obstructs the investigation, MOFCOM may make determinations based on the facts and information already obtained.

Article 13

During the investigation process, MOFCOM may engage in consultations with relevant foreign parties regarding the measures or acts under investigation.

Article 14

Where deemed necessary, MOFCOM may suspend an investigation.

Article 15

MOFCOM may terminate an investigation and publish a notice under any of the following circumstances:

The investigated country, region, organization, individual, or international organization has cancelled or adjusted the measure or act under investigation and has ceased causing harm or the threat of harm to China’s industrial and supply chain security;

The investigated country, region, organization, individual, or international organization has reached a solution with China;

Other circumstances under which MOFCOM determines that termination of the investigation is appropriate.

Article 16

Upon completion of an investigation, MOFCOM shall determine whether the investigated measures or acts have caused harm to China’s industrial and supply chain security and shall publish the determination.

Article 17

Based on the results of an investigation, MOFCOM may, together with relevant departments and in accordance with applicable procedures, take the following measures against countries, regions, or international organizations that implement or assist in acts harming China’s industrial and supply chain security:

Prohibit or restrict the import and export of goods and technologies, or international trade in services related to them;

Impose special charges or fees on activities related to them;

Pursuant to the Anti-Foreign Sanctions Law of the People’s Republic of China and the Provisions on Implementing the Anti-Foreign Sanctions Law, place organizations and individuals that directly or indirectly participate in formulating, deciding upon, or implementing the measures or acts described in Article 3(1) on a countermeasures list and impose countermeasures against them;

Other necessary measures.

Article 18

Based on the results of an investigation, MOFCOM may, together with relevant departments, take the following measures against foreign organizations and individuals that harm China’s industrial and supply chain security:

Prohibit or restrict their engagement in import and export activities related to China;

Prohibit or restrict their investment within China;

Prohibit or restrict organizations and individuals within China from engaging in transactions, cooperation, or other activities with them;

Prohibit or restrict the entry into China of relevant persons and means of transportation;

Revoke or restrict the qualifications of relevant persons to work, stay, or reside in China;

Other necessary measures.

The above measures may also apply to organizations that are actually controlled by, or whose establishment or operation involves, such foreign organizations or individuals.

Article 19

Where organizations or individuals within China fail to comply with measures imposed by MOFCOM pursuant to Articles 17 or 18 of these Measures, MOFCOM may order corrective action and, together with relevant departments, impose one or more of the following measures depending on the circumstances:

Prohibit or restrict their participation in government procurement, public tenders and bidding, import and export activities, international trade in services, and related activities;

Prohibit or restrict their receipt of data or personal information from abroad, or their provision of data or personal information to parties outside China;

Prohibit or restrict their departure from China, or their stay or residence within China;

Other necessary measures.

Article 20

Where deemed necessary, MOFCOM may adjust measures imposed pursuant to Articles 17, 18, and 19 of these Measures.

Article 21

These Measures shall be interpreted by MOFCOM.

Article 22

These Measures shall enter into force on the date of promulgation.

This new mechanism, set to take effect on July 1, 2026, establishes formal channels—including a dedicated hotline (010-12369) and an online portal—for individuals to report unauthorized exports of critical resources. A MOFCOM spokesperson framed the move as an “international common practice,” noting that many countries use public reporting to monitor export-control compliance.

MOFCOM Announcement No. 26 of 2026

Announcement on Further Improving the Reporting and Handling Mechanism for Violations of Export Controls on Strategic Mineral-Related Dual-Use Items

Issued by: Bureau of Industrial Security and Export Control

Date: June 24, 2026

To strengthen public oversight and combat violations of export controls on strategic mineral-related dual-use items, and in accordance with the Export Control Law of the People’s Republic of China, the Foreign Trade Law of the People’s Republic of China, and other relevant laws and regulations, the Ministry of Commerce (MOFCOM) has decided to further improve the reporting and handling mechanism for violations related to strategic mineral export controls.

The following matters are hereby announced:

I. Reportable Violations

Any organization or individual has the right to report suspected violations involving the export of strategic mineral-related dual-use items, including:

• Exporting controlled strategic mineral-related dual-use items without a license;

• Exporting beyond the scope, conditions, or validity period specified in an export license;

• Exporting items whose export is prohibited;

• Circumventing licensing requirements through modification, disassembly, or export as parts and components;

• Circumventing export controls through third-country transshipment;

• Illegally transferring controlled technologies through trade, licensing, investment, exchanges, gifts, exhibitions, testing, assistance, training, joint R&D, employment, consulting, or other means;

• Providing services such as agency, freight forwarding, logistics, customs declaration, e-commerce platforms, or financial services while knowingly facilitating export control violations;

• Instructing or assisting exporters, importers, or end users in evading export controls;

• Conducting transactions with importers or end users listed on a control list;

• Failing to apply for a license when exporters know or should know that goods, technologies, or services not specifically listed may pose risks described under Article 12 of the Export Control Law;

• Domestic importers or end users violating commitments made to MOFCOM;

• Accepting or committing to foreign government requests for inspections, visits, or on-site verification related to strategic mineral export controls without authorization;

• Other violations of laws and regulations governing export controls on strategic mineral-related dual-use items.

II. Reporting Channels

Reports may be submitted through:

1. The online reporting platform on the website of MOFCOM’s Bureau of Industrial Security and Export Control (https://aqygzj.mofcom.gov.cn);

2. The reporting hotline: +86-10-12369 (available on official working days from 8:30–11:30 and 14:00–17:00).

III. Reporting Requirements

Reporters are responsible for the accuracy of the information provided. Reports should generally include:

• Information about the reporter;

• Contact information;

• Information about the reported party;

• Details of the suspected violation;

• Whether the same matter has been reported to another authority;

• Any other information the reporter believes relevant.

Reports should be submitted in Chinese.

IV. Circumstances Under Which Reports Will Not Be Accepted

Reports will generally not be accepted if:

• The matter does not fall within the scope described above;

• Key information is missing and remains incomplete after MOFCOM requests supplementation;

• The matter has already been fully handled and is being repeatedly reported based on the same facts;

• Other circumstances where acceptance is deemed inappropriate.

V. Feedback and Rewards

For reports submitted under real names, MOFCOM will provide feedback on acceptance and handling status through appropriate channels. Individuals whose reports are verified may receive rewards in accordance with relevant regulations.

VI. Malicious Reporting

MOFCOM, together with relevant authorities, will carefully review suspected malicious reports and handle them in accordance with the law.

VII. Voluntary Disclosure

Exporters or other entities that discover they may have violated export control regulations are encouraged to voluntarily report the matter to MOFCOM. Voluntary disclosure may be considered a mitigating factor when determining penalties.

VIII. Confidentiality

Government agencies, organizations, and individuals involved in handling reports shall keep confidential any state secrets, commercial secrets, or personal information obtained during the process.

IX. Technology Export Violations

Reports involving suspected violations of strategic mineral-related technology export controls under the Regulations of the People’s Republic of China on the Administration of Technology Import and Export shall be handled with reference to this announcement.

This announcement shall take effect on July 1, 2026.

Ministry of Commerce of the People’s Republic of China
June 24, 2026

In a June 24 Q&A, the spokesperson grounded the policy in Article 31 of China’s Export Control Law—which gives “any organization or individual” the right to report suspected violations—and argued that borrowing from “the experience of various countries” would help keep strategic minerals from being “used for illegal purposes” while protecting the legitimate rights of compliant traders.

MOFCOM Spokesperson Answers Questions on Further Improving the Reporting and Handling Mechanism for Violations of Export Controls on Strategic Mineral-Related Dual-Use Items

Question: We have noticed that the Ministry of Commerce has issued an announcement further improving the reporting and handling mechanism for violations involving export controls on strategic mineral-related dual-use items. What is the rationale behind this move?

Answer: Article 31 of the Export Control Law of the People’s Republic of China provides that any organization or individual has the right to report suspected violations of export control regulations to the relevant state export control authorities. Strengthening the reporting and handling of violations involving strategic mineral-related dual-use items, and enhancing enforcement capabilities, is a basic requirement for improving China’s export control system and a normal measure to safeguard national security and interests.

At the same time, using reporting mechanisms to support enforcement of export control laws is a common international practice. Many countries have similar systems in place. Drawing on international experience and improving China’s reporting framework for strategic mineral-related dual-use items will help prevent such items from being used for illicit purposes, better contribute to global peace and security, and demonstrate China’s responsibilities as a major country.

In the past, we have already accumulated practical experience in handling reports of potential violations. It is therefore necessary to further clarify reporting channels, reporting requirements, and related procedures in order to improve the handling of leads and investigations.

Going forward, the relevant authorities will process reports in accordance with laws and regulations while safeguarding the legitimate rights and interests of compliant trading companies.

For the last few years, Washington has been aggressively building a similar apparatus, transforming whistleblowers into a front-line force for enforcing semiconductor export controls against China. People are witnessing the privatization of export-control enforcement. Whether you are a disgruntled employee in a Silicon Valley fab or a logistics coordinator in a Shenzhen warehouse, you are now a potential enforcement agent in a global rivalry.

Washington’s turn toward the informant economy was born out of a realization: traditional auditing is too slow for the speed of AI and advanced chips. The legislative foundation began to shift in late 2022 with the Anti-Money Laundering (AML) Whistleblower Improvement Act, which expanded reward programs to cover trade and economic sanctions violations. For the first time, tipsters could claim between 10% and 30% of collected fines that exceed $1 million.

However, a significant legal gap remained. While sanctions were covered, the Export Control Reform Act (ECRA)—the primary tool used to block the flow of semiconductors to China—lacked a dedicated whistleblower bounty.

The “Stop Stealing our Chips Act”, introduced by a bipartisan pair, Senators Mike Rounds (R-SD) and Mark Warner (D-VA), aims to close this loophole by creating a formal whistleblower incentive program within BIS. The Senate passed the bill unanimously on May 22, 2026, though as of late June, it has yet to be signed into law following its advancement in the House. If finalized, it would mirror the AML rewards, offering 10–30% of fines to those who expose semiconductor smuggling.

While Congress deliberates, the executive branch has already moved. In April 2023, BIS issued guidance incentivizing companies to report their competitors’ violations in exchange for “mitigation credit” on their own future liabilities. The Department of Justice followed suit with a Corporate Whistleblower Awards Pilot Program in August 2024, which was expanded in May 2025 to explicitly cover sanctions and trade fraud.

The results are already showing up in court. In December 2025, the DOJ dismantled “Operation Gatekeeper,” a smuggling ring that reportedly funneled $160 million worth of export-controlled Nvidia H100 and H200 GPUs to China. Months later, in March 2026, federal prosecutors charged a co-founder of Super Micro Computer for allegedly conspiring to divert $2.5 billion in Nvidia AI servers to the mainland. These cases are increasingly built on “implementation intelligence”—the kind of granular detail only an insider can provide.

In the similar way, China has spent the last two years constructing a parallel enforcement apparatus to police its own export controls on gallium, germanium, antimony, and rare earth elements.

The June 2026 MOFCOM announcement is the capstone of this effort. It builds on the 2020 Export Control Law, which already mandated whistleblower confidentiality, and the 2024 Dual-Use Items Export Control Regulations, which went into effect on December 1, 2024. These new rules even impose mandatory reporting obligations on service providers, like freight forwarders and banks, who “discover suspected violations” during their business operations.

In May 2025, China launched a multi-agency “Special Action to Crack Down on Strategic Mineral Smuggling,” involving the police, customs, and state security. The crackdown has already targeted high-profile figures. Lan Tianli, the Governor of Guangxi, was placed under investigation in May 2025; while official statements used broad language about “discipline violations,” contemporary reports linked the case to a family-run operation that allegedly misdirected mining rights for illicit export.

Beijing is also increasingly willing to leverage “competitive reporting.” In July 2025, Chinese media reported that a tip from the U.S. company Alpha Materials helped customs intercept a 300-ton shipment of disguised rare earth minerals, valued at $80 million, that was bound for the United States. Later that year, Shenzhen port officials intercepted 25 tons of NdFeB magnetic powder—a critical material for the F-35 fighter—that had been creatively disguised as “tile adhesive”. By formalizing the rewards for such “real-name” reporting in the June 2026 announcement, MOFCOM is signaling that it wants more of these insider-driven busts.

According to Lianhe Zaobao—citing Japan’s Asahi Shimbun and Kyodo News—Chinese authorities detained two Japanese nationals in the northeastern city of Dalian, Liaoning Province, in May 2026 on suspicion of “smuggling goods prohibited from import or export by the state”. One of them, reportedly a Japanese male employee of a large local Japanese-funded electronics company, was accused of trying to ship products made with export-controlled rare-earth elements out of China. Japanese Chief Cabinet Secretary Minoru Kihara disclosed the cases on June 24, 2026, saying the two had been detained on May 18 and May 25, with Chinese customs notifying Japan’s diplomatic missions the day after each detention.

The privatization of export controls changes the risk calculus for every company operating in the U.S.-China corridor. In the past, compliance was a game played against the regulator; today, it is a game played against your own internal data and your own workforce.

Specifically, the 10 entities were added to MOFCOM’s export control list, meaning they can no longer receive Chinese-origin controlled items. Separately, the Ministry of Finance imposed government procurement restrictions on 46 U.S. companies, effectively excluding them from China’s government procurement market.

MOFCOM Announcement No. 23 of 2026
Decision on Adding 10 U.S. Entities to the Export Control List

Issuing Authority: Bureau of Industrial Security and Export Control, Ministry of Commerce

Document No.: MOFCOM Announcement No. 23 of 2026

Date of Issuance: June 22, 2026

In accordance with the relevant provisions of the Export Control Law of the People’s Republic of China, the Regulations of the People’s Republic of China on Export Control of Dual-Use Items, and other applicable laws and regulations, and for the purposes of safeguarding national security and interests and fulfilling international obligations such as non-proliferation, it has been decided to add the following 10 U.S. entities to the Export Control List (see Annex) and to impose the following measures:

I. Export operators are prohibited from exporting dual-use items to the 10 entities listed below. Any organization or individual in any country or region is prohibited from transferring or providing dual-use items originating in China to these entities. Any ongoing export activities related to these entities shall be immediately suspended.

II. Where export is genuinely necessary under special circumstances, the exporter shall apply to the Ministry of Commerce for approval.

This Announcement shall take effect immediately upon its issuance.

Annex: Export Control List (June 22, 2026)

Ministry of Commerce of the People’s Republic of China

June 22, 2026

Annex

Export Control List

(June 22, 2026)

1. Aveox, Inc.

Address: 2265A Ward Ave., Simi Valley, CA 93065, USA

Also Known As: AVEOX

2. Red Cat Holdings, Inc.

Address: 2800 S West Temple St., Unit 2, South Salt Lake, UT 84115, USA

Also Known As: Red Cat

3. Teal Drones, Inc.

Address: 2800 S West Temple St., Unit 2, South Salt Lake, UT 84115, USA

Also Known As: Teal Drones, iDrone

4. IMSAR, LLC

Address: 940 S 2000 W #140, Springville, UT 84663, USA

Also Known As: IMSAR

5. Jaia Robotics, Inc.

Address: 22 Burnside St, Bristol, RI 02809, USA

Also Known As: Jaia Robotics

6. Ball Aerospace & Technologies Corp.

Address: 10 Longs Peak Drive, Broomfield, CO 80301, USA

Also Known As: Ball Aerospace; Space & Mission Systems business of BAE Systems

7. Oshkosh Defense, LLC

Address: 2307 Oregon Street, Oshkosh, WI 54902, USA

Also Known As: Oshkosh Defense

8. L3Harris Maritime Services, Inc.

Address: 3835 E Princess Anne Rd, Norfolk, VA 23502, USA

Also Known As: L3Harris Maritime

9. MP Materials Corp.

Address: 1700 S Pavilion Center Drive, Eighth Floor, Las Vegas, NV 89135, USA

Also Known As: MP Materials

10. USA Rare Earth, Inc.

Address: 100 W Airport Rd, Stillwater, OK 74075, USA

Also Known As: USAR, USARE

Notice on Taking Relevant Measures Against Certain U.S. Companies in Government Procurement Activities

Caiku [2026] No. 10

To all central government budget entities, the finance departments (bureaus) of all provinces, autonomous regions, municipalities directly under the Central Government, and cities specifically designated in the state plan, and the Finance Bureau of the Xinjiang Production and Construction Corps:

In accordance with relevant laws and regulations, and upon approval, it has been decided to take relevant measures against 46 U.S. companies in government procurement activities. The details are hereby notified as follows:

I. In government procurement activities, procuring entities shall not purchase products manufactured by the 46 U.S. companies listed in the Annex (excluding U.S.-invested enterprises established in China).

II. This Notice shall take effect as of the date of issuance.

Annex: List of 46 U.S. Companies

Ministry of Finance of the People’s Republic of China

Caiku [2026] No. 10

On June 8, the U.S. Department of War updated its Section 1260H list. The updated list includes 17 Chinese companies, among them Alibaba, Baidu, and BYD. The Pentagon also made clear that additional companies could be added in the future. At the same time, the U.S. government emphasized that inclusion on the 1260H list does not automatically trigger sanctions, although it could pave the way for future restrictions under other legal authorities.

The legal basis for the list is Section 1260H of the FY2021 National Defense Authorization Act (NDAA). Originally, the list was designed to identify companies that the U.S. government believes are linked to China’s military and that operate, directly or indirectly, in the United States. It is not the Treasury Department’s SDN list, so it does not automatically freeze assets. It is not the Commerce Department’s Entity List, so it does not automatically block exports. Nor is it the NS-CMIC list, which restricts U.S. investment in certain Chinese companies. For many years, the market largely viewed 1260H as a naming-and-shaming exercise and a reputational risk label.

That changed a bit with Section 805 of the FY2024 NDAA.

Starting on June 30, 2026, the DoW is prohibited from directly entering into or renewing contracts for goods or services with companies on the 1260H list, as well as entities they control. This is the first clear and tangible legal consequence of being listed. In practical terms, companies on the list will be shut out of the Pentagon’s supplier ecosystem.

The restriction is not limited to traditional defense products. It can affect cloud services, software, AI capabilities, data-processing services, robotics, batteries, memory chips, autonomous-driving technologies, and other digital products and services if the ultimate customer is the DoW or if the contract ultimately flows into the DoW procurement system.

The restrictions become even broader one year later. Beginning June 30, 2027, the Pentagon will not only be barred from buying directly from listed companies, but also from procuring their products, components, or services indirectly through prime contractors, subcontractors, or deeper tiers of the supply chain. For large technology companies, the bigger challenge is often not whether they sell directly to the DoD, but whether their technology, components, compute, storage, batteries, or software are embedded in products supplied by others.

The importance of the 1260H list also lies in its spillover effects. Once a company is placed on the list, it effectively enters the U.S. national security system’s watchlist, lowering the political and bureaucratic barriers for future restrictions.

From a commercial perspective, the bigger concern is often overcompliance. Many multinational companies treat a 1260H designation as a red flag. Rather than spending time analyzing whether cooperation remains legally permissible, they often choose the safer route: reduce exposure and look for alternative suppliers. As a result, the practical impact of the list can extend well beyond its formal legal consequences.

So when I saw such a large-scale expansion of the 1260H list, my initial view was that Beijing probably wouldn’t just let this slide. I expected a strong public response, followed by some limited countermeasures. Three reasons:

1/ The scope of the list has expanded significantly. Traditionally, 1260H focused on defence-related SOEs. This time it reaches into cloud services, EVs, biotech, robotics, and drones—all sectors at the core of China’s “new productive forces” strategy. Beijing is likely to see this as an attempt to constrain China’s future technological advantages, which makes it politically difficult to shrug off.

2/ The direct impact is manageable, but it’s not just a reputational issue. NDAA Section 805 has linked the list to DoD procurement restrictions, and the risk of over-compliance by international business partners could amplify the indirect effects.

3/ This is arguably the first major and politically significant escalation of U.S. restrictions on China since the Busan summit. It also lays the groundwork for tougher measures down the road, including investment restrictions and supply-chain exclusions. Beijing may feel that failing to respond would only encourage more of the same. At the same time, a calibrated response would not necessarily derail broader stability in the relationship.

This interpretation was reinforced by the subsequent responses from both MOFCOM and the Ministry of Foreign Affairs. Both signaled that China would retaliate, but MOFCOM's statement was notably sharper in tone and, by recent standards, somewhat unusual.

MOFCOM Spokesperson Responds to U.S. Designation of Chinese Companies on the “Chinese Military Companies” List

Question: On June 8 (U.S. time), the Department of Defense added several Chinese companies to its “Chinese Military Companies” list. What is MOFCOM’s response?

Answer: We have noted the development. China is strongly dissatisfied with and firmly opposed to this action.

The U.S. has disregarded the consensus reached by the two heads of state in Beijing, ignored the broader interests of China-U.S. economic and trade relations, continuously overstretched the concept of national security, and abused state power to suppress Chinese companies without justification. These actions seriously undermine the international economic and trade order, disrupt the stability of global industrial and supply chains, and harm the legitimate rights and interests of Chinese enterprises.

China urges the United States to immediately stop its wrongful practices, revoke the relevant measures, and return to the right path of building a constructive and strategically stable China-U.S. relationship. The U.S. should provide Chinese companies with fair, equitable, and non-discriminatory treatment.

Otherwise, China will take firm and forceful countermeasures, and the consequences and responsibility arising from them will rest entirely with the U.S. side.

MOFA Q&A

Question: The United States has added a number of Chinese technology companies, including Alibaba, Tencent, and BYD, to its list of companies allegedly supporting China’s military. What is China’s response?

Lin Jian: China has consistently and firmly opposed the U.S. practice of overstretching the concept of national security, creating various discriminatory lists, and unjustifiably suppressing Chinese companies.

We urge the United States to correct its mistakes and stop its unreasonable actions against Chinese enterprises.

China will take necessary measures to firmly safeguard the legitimate rights and interests of Chinese companies.

But at that time, I still thought the odds of such a response were probably only around 30%. A 70% chance that Beijing limits itself to criticism and does not retaliate. Several reasons：

Historically, even when those high-profile companies like Tencent and CATL were added to the 1260H list last year, Beijing’s reaction amounted to strong criticism and promises to defend the rights of Chinese companies, but no visible follow-up retaliation ever materialised.

This designation is not particularly serious, as we explained above. Since the Busan understanding, Washington has already taken several steps that could be seen as changing the status quo. The FCC has kept up pressure on Chinese firms, the Treasury has occasionally added companies to the SDN list, and Beijing has largely chosen not to respond.

In addition, in some of these sectors where China believes it has a technological edge, Beijing may not mind a degree of decoupling from the U.S. There are a couple of recent examples that may point in that direction.

But Beijing retaliated anyway, even though the response was largely symbolic.

The targets remain concentrated in the defense and rare-earth sectors, many of which have already been sanctioned previously. Beijing did not expand the measures to broader categories of U.S. companies. At the same time, the Ministry of Finance’s government procurement restrictions were carefully limited to U.S.-origin products from 46 American firms, explicitly excluding the products of U.S.-invested companies operating and manufacturing in China. In other words, products made in China by those firms can still participate in government procurement.

On the specific measures:

1. Export-control restrictions on 10 entities

Most of the newly listed entities are defense-oriented contractors whose business is already heavily tied to Pentagon programs and which generally do little or no business with China. The main impact falls on their supply chains, particularly where rare earth magnets, gallium, germanium, or semiconductor-related inputs are involved. However, this is not a new vulnerability.

The most notable additions are MP Materials and USA Rare Earth. In practice, obtaining Chinese heavy rare earth feedstock or separation technology was already highly unlikely for these companies, so the practical impact may be limited even if the political signal is significant.

2. Government procurement restrictions on 46 U.S. firms

The Ministry of Finance’s decision to bar 46 U.S. companies from government procurement is more noteworthy. China used a similar tool against EU medical device firms in 2025, but this is the first time it has been applied to American companies.

The list covers a broad cross-section of the U.S. defense industrial base. However, because the United States has maintained a comprehensive arms embargo on China since 1989, most of these companies were never realistically selling products to Chinese government agencies or the PLA in the first place. A few firms with commercial business lines, such as Oceaneering or Teledyne/FLIR, have some exposure to China, but even there government procurement represents a relatively small share.

As a result, the direct economic impact is limited. The more important message is one of reciprocity: if the U.S. government prohibits the Pentagon and its contractors from purchasing products from designated Chinese companies under the 1260H framework, China can likewise prohibit its own government agencies from purchasing products from designated U.S. companies.

One thing that is notable is that most of these companies are somewhat related to Taiwan or US arms sales to Taiwan.

A commentary published by China’s Global Times argued that many of the targeted companies have long been involved in U.S.-Taiwan defense cooperation. Citing Taiwan’s Central News Agency, it noted that in September 2024, more than 20 U.S. drone, counter-drone, and unmanned maritime systems companies visited Taiwan as part of an AIT-organized trade delegation, including Jaia Robotics. The commentary also highlighted that L3Harris Maritime Services and Red Cat participated in the Taipei Aerospace & Defence Technology Exhibition last year.

The commentary further quoted military analyst Zhang Junshe, who noted that

China had already sanctioned L3Harris Maritime Services, Teal Drones, and Red Cat through Foreign Ministry measures in December 2025. The fact that these companies are now being targeted again by MOFCOM reflects Beijing’s view that they have played a particularly significant role in supporting Taiwan-related military activities. The combined measures are intended to cut off trade, technology cooperation, and technology transfers involving dual-use items, thereby reducing external support for pro-independence forces in Taiwan.

Regarding MP Materials and USA Rare Earth, the commentary pointed to Bloomberg reporting that Washington is actively building alternative critical minerals supply chains to support advanced manufacturing and defense industries. It argued that the latest Chinese measures are aimed directly at key nodes in that supply chain. The commentary also noted growing U.S.-Taiwan cooperation on critical minerals, including commitments made during the sixth U.S.-Taiwan Economic Prosperity Partnership Dialogue earlier this year.

Zhang Junshe argued that:

Rare earths are essential inputs for military platforms such as fighter aircraft, warships, missiles, and military drones. Since the United States uses rare earths to manufacture weapons that are subsequently sold to Taiwan, placing MP Materials and USA Rare Earth under export-control restrictions is, in Beijing’s view, a way to weaken the material foundation of future U.S. arms sales to Taiwan.

Li Haidong, a professor at the China Foreign Affairs University, was quoted as saying that

China would seek to “maximally weaken and strike” U.S. companies that play a major role in arming Taiwan.

My main takeaways are:

First, Beijing appears to have a fairly accurate understanding of what the 1260H list actually is and didn’t overreact. Today’s measures were likely not aimed solely at the latest 1260H designations. They appear to be a broader response that also reflects accumulated frustration over recent SDN listings, FCC restrictions, and other U.S. actions.

Second, that restraint should not be mistaken for passivity. China’s longstanding approach of “fighting when necessary in order to preserve stability” remains intact. If the U.S. continues to roll out additional restrictive measures, Beijing is likely to respond in kind.

Third, the remarkable addition of government procurement restrictions seems to send a clear warning to Washington: many non-defence U.S. companies operating in China still have commercial interests tied to government procurement, and those interests could become targets in future rounds of escalation.