含羞草社区

In 2025, during the Chinese New Year, China’s AI model, DeepSeek, gained global attention by surpassing ChatGPT to become the most downloaded app on the Apple application store in both China and the US. This indicates China’s progress in AI and is in line with its national strategy.

In 2017, the State Council introduced the New Generation AI Development Plan with a “three step” legislation roadmap. By 2020, the goal was to match global AI standards. By 2025, China aimed for major breakthroughs in AI theory and the world-leading technology and applications for industrial and social progress. Looking ahead to 2030, China aspires to be a global leader in AI.

This article examines China’s AI regulation, focusing on its legislative framework, compliance, law enforcement and future prospects.

Legislative framework

China’s AI regulation is built on a multi-level framework of laws and regulations, covering data compliance, algorithm compliance, cybersecurity and ethics.

Data compliance. China’s data compliance is guided by fundamental laws and regulations, including the:

1. 1. Personal Information Protection Law (2021);
   2. Data Security Law (2021); and
   3. Regulation on Network Data Security Management (NDSM), which is set to take effect in 2025.

Cybersecurity. The regulatory framework for cybersecurity is based on laws and regulations, such as the:

1. 1. Cybersecurity Law (CSL) (2017); and the
   2. NDSM, effective in 2025.

Ethical review. For ethical review, the system is based on the following fundamental laws and regulatory documents:

1. 1. Law on the Progress of Science and Technology (revised in 2022); and
   2. Measures for Science and Technology Ethics Review (Trial) (2023).

Algorithm compliance. Under algorithm compliance, there are both departmental and local regulations. The following frameworks apply under departmental regulations:

1. 1. Provisions on the Security Assessment of Internet-based Information Services with Attribute of Public Opinions or Capable of Social Mobilisation (2018);
   2. Provisional Measures for the Administration of Generative Artificial Intelligence Services (GAI measures, effective in 2023);
   3. Administrative Provisions on Deep Synthesis of Internet-based Information Services (2023); and
   4. Administrative Provisions on Algorithm Recommendation for Internet Information Services (2023).

Meanwhile, the following frameworks are established under local regulations:

1. 1. Regulations on Promoting the Development of Artificial Intelligence Industry in Shanghai (2022);
   2. Regulations on the Promotion of Artificial Intelligence Industry in the Shenzhen Special Economic Zone (2022);
   3. Fujian Province Artificial Intelligence Industry Development Project Management Measures (2024); and
   4. Guiding Opinions on Accelerating the Development of the Artificial Intelligence Industry in Zhejiang Province (2023).

Aside from the departmental and local regulations, China’s primary regulations on algorithm compliance include the Regulations on the Identification of Artificial Intelligence-Generated Synthetic Content (Draft for Comment).

To further implement the above-mentioned laws and regulations, China has also successively introduced a series of national and industry standards and specifications. For all industries, they include:

1. 1. Code of Ethics for the New Generation Artificial Intelligence (2021);
   2. Basic Security Requirements for Generative Artificial Intelligence Services (2024); and
   3. Cybersecurity Technology – Labelling Method for Content Generated by Artificial Intelligence (Draft for Comment).

For the medical industry, they include:

1. 1. Regulatory Rules for Internet-based Diagnosis and Treatment (Trial) (2022);
   2. Guiding Principles for the Classification and Definition of AI-based Medical Software Products (2021); and
   3. Guiding Principles for the Review of Artificial Intelligence Medical Device Registrations (2022).

Finally, for the intelligent-connected vehicles industry, Good Practice for the Administration of Road Tests and Demonstrative Application of Intelligent and Connected Vehicles (Trial) (2021) serves as the industry standard.

Compliance requirements

The GAI measures focus on generative AI service providers, imposing legal obligations like algorithms, content, IP, training corpus and data annotation compliance.

Algorithm filing and security assessment. AI services in China with “public opinion or social mobilisation capabilities” must file their algorithm mechanisms with the Cyberspace Administration of China. This includes text, pictures, voice and video-generation AI services. Services without these attributes are exempt from filing. Non-compliant AI providers may face severe penalties, from service suspension to criminal liability. All AI services must also pass a security assessment before filing. Failed services cannot be launched, and existing ones need rectification or removal. These regulations ensure AI services operate safely in China.

Content marking. AI service providers must label generated or edited content using non-intrusive identifiers and retain log information as required by laws. Deep-synthesis services that may cause public confusion need to clearly label content in an appropriate location. No one is allowed to use technical means to delete, alter or conceal these labels. These regulations promote transparency and prevent misinformation in AI-generated content.

Science and technology ethics review. AI providers conducting research in sensitive areas, including algorithm models, applications and systems that influence public opinion and societal awareness, must establish a science and technology ethics (review) committee. They must also carry out ethical risk assessments and review work in compliance with the law.

Data compliance. The data compliance obligations include:

1. 1. Training data. Obligations include the corpus source security obligations, corpus content security obligations, corpus annotation security obligations, and the establishment of compliant lexicons and question banks;
   2. Input data. Providers are not allowed to collect unnecessary personal information, illegally retain input information and usage records that can identify the user’s identity, or illegally provide the input information and usage records of users to others. Deep synthesis providers shall obtain their separate consent; and
   3. Output content. Generative AI must comply with laws, respect for morality and ethics, and uphold socialist values. Content that threatens the state and society, or promotes harmful ideologies and false information, must be avoided.

Law enforcement practice

While AI development has brought great convenience, it has also raised challenges that exceed existing legal boundaries, prompting China to accelerate legislative improvements.

Judicial practice. AI-related judicial cases mainly involve personality and IP rights infringement.

1. 1. Infringement of personality rights. In 2021, the Beijing Internet Court handled an AI-related personality rights case in which a public figure sued over an app’s “AI companions”. The court found the defendant liable, as its design and algorithms promoted user-generated content. The case was later recognised as typical in 2022. In 2024, the court issued its first ruling on AI-generated voice infringement, ordering compensation. That same year, it heard two “AI face-swapping” cases, ruling that the unauthorised use of models’ images for templates violated the plaintiffs’ personal information rights, requiring compensation and an apology.
   2. Infringement of intellectual property rights. In 2023, the Beijing Internet Court ruled on China’s first AI-generated painting copyright case, protecting relevant works. The Guangzhou Internet Court also handled an AI-related painting copyright infringe-ment case, ordering comp-ensation. In 2024, China’s first AI-generated audio-visual work infringe-ment case entered pre-trial proceedings. The plaintiff accused the defendant of infringement and demanded comp-ensation. The case is ongoing.

Administrative supervisory practice. AI-related administrative penalties have mainly revolved around enterprise qualifications and consumer rights protection. So far, no cases have involved penalties for personal information protection, data security or network security violations.

Prospects

Tech development and industry use. A series of domestic AI tools such as DeepSeek, Doubao and Kimi have emerged. DeepSeek innovates algorithms by optimising architecture to boost computing power efficiency, challenging the traditional power-centred model. Thus, China’s AI and related technologies are set for faster development and have broad prospects.

AI law formulation. Although no major fundamental AI legislation was passed in 2024, the introduction of detailed regulations and active participation in shaping international rules indicate steady legislative efforts. In the near future, China may introduce the Artificial Intelligence Law (basic law). This bill was included in the legislative plan as early as 2023, and listed as a preparatory review project in the State Council’s legislative agenda in 2024.

AI legislation: Efficiency v safety

China has built a highly systematic AI regulatory framework through “legislation first, ethical guidance and classified governance”. The key for future legislation and practice is balancing technology innovation with risk control. The rise of domestic AI tools like DeepSeek will speed up the process of Artificial Intelligence Law, and China may offer a “Chinese model” for global AI governance, balancing efficiency and safety.

BEIJING DACHENG LAW OFFICES
Floor 9/24/25, No.100 Century Avenue
Shanghai, China
Tel: +86 21 5878 5888