China’s AI Law: Recent Developments and Legislative Signals

On June 17, 2025, Legal Daily, one of China’s most authoritative legal newspapers, published a front-page report titled “Multiple NPC Representatives Propose Drafting an Artificial Intelligence Law: Building a Comprehensive and Scientific Legal Framework for AI.” The article was reposted the same day on the official website of the National People’s Congress (NPC), China’s highest legislative body.

Based on the report, during this year’s NPC session, Quan Heng (权衡) and several other deputies submitted proposals on enacting China’s AI Law. The proposal emphasized leveraging the guiding and enabling role of legislation to formulate a comprehensive and systematic legal framework for AI, aiming to support the building of a strong, modern nation and to advance Chinese-style modernization in the new era.

Quan Heng is currently the Party Secretary of the Shanghai Academy of Social Sciences and a member of the Shanghai delegation to the National People’s Congress (NPC). He has long been involved in policy research and legislative proposals related to regional economic and social development, with a particular focus on Yangtze River Delta integration, high-quality development, and reform and opening-up. He has been an active voice in putting forward legislative and policy recommendations during NPC sessions.

While the full text of his proposal—like others submitted during the session—has not been made public, Legal Daily did disclose several key points from it:

The proposal argues that, at present, the development of the AI industry urgently requires legal safeguards. The development of AI technology and industry brings a host of new risks, which must be addressed through dedicated legislation that regulates dynamic risk classification and fault-tolerance mechanisms. Existing laws are limited in scope. Laws such as the Data Security Law and the Personal Information Protection Law, though related to AI, mainly focus on the regulation of single elements. Local regulations, such as the Regulations on Promoting the Development of the AI Industry in Shanghai, are relatively low in legal hierarchy and struggle to address the complexity of AI technology applications that span across regions. Therefore, it is deemed highly necessary to draft a national-level AI Law.

The proposal recommends that AI legislation should focus on several key areas:

• In terms of promoting innovation and industrial development: a series of incentive policies should be introduced, including increased financial investment in AI R&D and applications, the formulation of financial policies conducive to AI industry development, the strengthening of talent cultivation and recruitment in the AI field, and the clarification of development plans for the AI industry.

• In terms of risk classification and regulatory measures: AI applications should be categorized into prohibited practices, high-risk AI systems under strict regulation, AI systems with transparency risks, and low-risk AI systems. Regulatory measures should be formulated according to the different types of risks.

• In terms of social impact and ethical considerations: it is necessary to require assessments of the social impact of AI technologies, to formulate ethical standards for AI technologies, and to define the moral boundaries and codes of conduct for the use of such technologies.

• In terms of international competition and cooperation: attention should be paid to the global competitive landscape. Corresponding competition strategies and development plans should be proposed, while also emphasizing the importance of international cooperation and exchange, including putting forward specific measures to strengthen such cooperation.

• In terms of legal liability and regulatory mechanisms: the law should clarify legal responsibilities related to AI technologies, including the definition of illegal acts, assumption of legal liability, compensation mechanisms, and regulatory mechanisms.

The proposal concludes that the identification and refinement of these key areas will contribute to building a comprehensive, scientific, and rational legal system for AI, thereby providing strong legal guarantees for the healthy development of AI technology.

Earlier, on May 28, Zhang Linghan (张凌寒)—a prominent expert in AI law and a professor at the Institute for Data Law and Governance at China University of Political Science and Law—also published a commentary in Legal Daily.

In the article, Zhang pointed out:

“Not long ago, the 2025 Legislative Work Plan of the State Council was released. Unlike the 2023 and 2024 plans, this year’s version no longer lists the draft AI Law as a preparatory item for submission to the NPC Standing Committee. Instead, the wording has been revised to 'advance legislation for the sound development of AI.' A corresponding statement appeared in the 2025 Legislative Work Plan of the NPC Standing Committee, released on the same day: 'Legislative items concerning the sound development of AI shall be researched and drafted promptly by relevant departments, and deliberation shall be arranged as appropriate.' The two plans echo each other.”

Zhang went on to raise a central question:

“This change has drawn attention and discussion from academia and the tech industry. So, how should we understand the shift in language?”

According to Zhang, the recent adjustment in the official language around AI legislation does not signal a decline in its importance or urgency. Rather, it reflects a more systematic and strategic approach to advancing the legal framework. Against the backdrop of the state’s continued emphasis on AI legislation, this shift in wording is intended to better coordinate top-level, specialized AI laws with related elemental and sectoral laws, and to promote the development of a multi-layered, comprehensive legal system covering data, algorithms, and applications. The goal is to provide a more robust rule-of-law foundation for the healthy, secure, and high-quality development of AI in China.

A legislative work plan, as a comprehensive arrangement formulated by a body with legislative authority for a specific period of lawmaking activity, must both determine legislative items based on the needs of economic and social development and the advancement of rule of law, and make adjustments in response to evolving circumstances and tasks. In practice, adjustments to legislative projects are the norm rather than the exception. For example, the drafting of the Stamp Tax Law and the amendment of the Company Law were not continuously listed in annual legislative plans, yet these laws were eventually enacted. Similarly, laws such as the Family Education Promotion Law and the Noise Pollution Prevention and Control Law were adopted under titles different from those originally listed in legislative plans. Legislative planning must strike a balance between flexibility for timely adjustments and persistence over the long term. For instance, regulations such as the Regulation on the Protection of Minors in Cyberspace and the Implementation Regulations for the Law on the Protection of Consumer Rights and Interests of the PRC were repeatedly added to and removed from legislative plans before eventually being adopted.

The reason AI legislation has attracted so much attention lies in the strategic significance of AI as a core technology driving the new round of scientific and industrial transformation. Reviewing the development of AI legislation in China:

• In 2017, the New Generation Artificial Intelligence Development Plan proposed establishing a preliminary framework of AI-related laws, ethical norms, and policy systems by 2025.

• In 2021, the Plan for Building Rule of Law in China (2020–2025) explicitly called for timely follow-up research on legal systems related to AI.

• In July 2024, the Third Plenary Session of the 20th CPC Central Committee adopted the Decision on Further Deepening Reform Comprehensively and Advancing Chinese-Style Modernization, which called for improving the development and governance mechanisms for generative AI and establishing a regulatory system for AI safety.

All of this demonstrates China’s consistent approach to safeguarding the healthy development of new industries and models through sound laws and good governance. The evolution in wording also reflects the overall acceleration of AI legislative efforts. Therefore, in the author’s view, the adjustment in the formulation concerning AI legislation does not reflect any change in the perceived necessity or urgency of such legislation, but instead stems from deeper strategic considerations.

From a procedural standpoint, the coordination between the State Council’s annual legislative plan and that of the Standing Committee of the National People’s Congress is not only in line with the Legislation Law’s requirement for aligned legislative planning, but also enhances the systematic, holistic, and coordinated development of AI legislation, thereby providing a more scientific, stable, and comprehensive legal foundation for AI’s sound development.

From the perspective of substance, the revised wording reflects a strategy of promoting AI legal system development in a more structured and integrated manner. In recent years, China has made steady progress in AI-related legislation and has begun to build a multi-level legal framework that includes laws, administrative regulations, departmental rules, and local regulations, covering multiple dimensions such as data, algorithms, and applications. Given that AI is a “civilizational-level technology” that affects both technological progress and industrial transformation, the legislative layout must encompass not only top-level, specialized AI laws, but also a broader range of elemental laws and sector-specific laws. Thus, the use of the phrase “advance legislation for the sound development of AI” is intended to support high-quality development and high-level security in a coordinated manner and to promote a comprehensive and systematic approach to building the AI legal system.

From the perspective of legislative trends, in the context of intensifying global technological competition, top-level legislation for AI remains a top priority in the construction of China’s AI legal system. Although preliminary steps have been taken to regulate technologies such as generative AI, fundamental legal issues—such as the legal status of AI, its governance structures, and its liability mechanisms—still require a unified, comprehensive law to provide consistent regulation. At the same time, as top-level legislation progresses, attention must also be paid to the coordination between supporting elemental and sectoral laws, in order to effectively guide China’s AI technology and industry along a rule-of-law path toward healthy, sustainable, and high-quality development.

Against the broader backdrop of U.S.-China competition over AI, China's approach to AI legislation has never been purely a legal or technical matter. From the outset, it has carried strong geopolitical undertones and is closely intertwined with China’s efforts to position itself in the global AI landscape.

In Washington, the absence of a comprehensive national AI law at the NPC level in China has already been cited by some U.S. companies and policy stakeholders as a reason to caution against imposing overly strict regulatory measures in the United States. Recently, there has been intense debate in Congress over a clause in the proposed “big, beautiful bill” that would impose a ten-year moratorium on state-level AI regulations. How this will ultimately unfold remains uncertain.

In my view, the adjusted language in the 2025 Legislative Work Plan of the NPC Standing Committee clearly continues to reflect the Chinese government’s strong focus on AI development. But I also agree with Professor Zhang: this does not mean China intends to abandon regulation.

In fact, China has already begun building a regulatory framework around generative AI and is gradually exploring how to enforce it. Today, the Shanghai Cyberspace Administration issued a public notice announcing penalties against a number of generative AI service platforms that had refused to make required corrections. This marks the first publicly disclosed case in which a Chinese regulator has formally penalized domestic AI firms for violating laws such as the Interim Measures for the Management of Generative Artificial Intelligence Services and the Regulations on the Management of Deep Synthesis Internet Information Services.

“Operation Bright Sword on the Huangpu”: Shanghai Cyberspace Administration Penalizes AI Service Providers for Non-Compliance

In a recent notice titled “Operation Bright Sword on the Huangpu: Protecting Personal Information Rights and Preventing the Generation of Harmful Content – Shanghai Cyberspace Administration Penalizes Non-Compliant Generative AI Service Websites”, the Shanghai Cyberspace Administration announced formal penalties against several companies that failed to comply with regulatory requirements on generative AI services.

The enforcement action is part of the “Bright Sword Huangpu · 2025” special campaign jointly led by the Cyberspace Administration, the Shanghai Administration for Market Regulation, and other relevant industry regulators. The campaign focuses on addressing compliance risks and safety concerns in the generative AI sector, with an emphasis on curbing misuse of AI technologies.

Key findings and regulatory actions:

Authorities found that some generative AI service providers had:

• Failed to conduct legally required security assessments;

• Failed to implement safeguards to prevent the generation of prohibited content;

• Failed to impose technical restrictions to prevent the misuse of generative AI functions.

Although companies were previously ordered to take these services offline until compliant, inspections revealed that three companies had resumed such services without approval or adequate security measures. These companies were formally interviewed, ordered to fully rectify the violations, and have now been placed under official investigation.

Typical violations identified include:

1. Generation of content infringing on personal information rights
   Some providers failed to comply with the Personal Information Protection Law and the Regulations on Deep Synthesis of Internet Information Services, such as by not informing or obtaining separate consent from individuals before processing their biometric data.

   Example: One website cloned a user’s voiceprint to generate synthetic speech without the person’s consent, posing a clear risk in handling biometric data.

2. Generation of content infringing on reputation, privacy, and other legal rights
   Some platforms did not implement necessary safeguards to prevent users from generating harmful content.

   Examples: One site produced fake images of public figures based on user input; another allowed users to generate "doxxing tutorials" or step-by-step guides to exposing private information.

3. Generation of prohibited content such as pornography or violence
   Some services failed to act in accordance with the Cybersecurity Law and the Interim Measures for the Management of Generative AI Services by not taking timely actions (e.g., blocking, deleting) when illegal content appeared.

   Examples: Platforms allowed the generation of pornographic or violent images, or text content with explicit sexual implications.

But regulation can take many forms. Just as there is ongoing debate in the U.S. between a unified federal approach and a more fragmented state-level model, China’s legislators, policymakers, intellectuals, and industry leaders have also engaged in extensive discussions over the past few years about what constitutes the right regulatory model for AI in the Chinese context.

Given the current domestic and international landscape China faces in the AI domain, I would be genuinely surprised if the National People’s Congress were to adopt a comprehensive regulatory framework resembling the EU AI Act. Clearly, there are still intermediate forms between a full-fledged "China AI Law" and the current patchwork of regulatory measures.

China’s policymakers have the strategic judgment and flexibility needed to strike a balance between development and security in the realm of AI regulation.