China has an extensive set of binding AI rules without a single comprehensive statute. Instead the Cyberspace Administration of China (CAC) and partner agencies issue targeted rules, each aimed at a specific problem: recommendation algorithms, deep synthesis, generative AI services, and content labeling. Public-facing generative services must pass security assessment and file with regulators, and AI-generated content must carry both visible and embedded labels. A comprehensive national AI law has been on the agenda since 2023 but remains a proposal.
Key points
- The Interim Measures for Generative AI Services took effect August 15, 2023, China's first rule written specifically for generative AI.
- The AI content-labeling Measures took effect September 1, 2025, requiring explicit (visible) and implicit (metadata) labels.
- Deep-synthesis rules require consent before recreating a real person's face or voice.
- Providers whose services can shape public opinion must file algorithms and pass a security assessment.
- A comprehensive national AI law has been on the legislative agenda since 2023 and has not been enacted as of October 2026; the State Council's 2026 legislative work plan of May 11, 2026 again calls for accelerating comprehensive AI legislation, and the national informatization plan for 2026 to 2030, published October 23, 2026, calls for advancing foundational legislation on AI.
- Labeling enforcement has run since November 2025; on September 15, 2026 the CAC published cases in which a generative AI mini program was ordered offline for missing explicit and implicit labels and skipping security assessment, and an API relay service was sanctioned for the same assessment failure.
Laws and instruments
Interim Measures for the Management of Generative AI Services
CAC and six agencies; effective August 15, 2023
China's first generative-AI regulation. Requires lawful training data, content moderation with provider responsibility, protection of personal information, and adherence to content rules.
Effective: August 15, 2023
Measures for Labeling AI-Generated and Synthetic Content
CAC, effective September 1, 2025; standard GB 45438-2025
Requires explicit labels users can see and implicit labels embedded in file metadata on AI-generated content, backed by a mandatory national standard.
Effective: September 1, 2025
Deep Synthesis Provisions
Provisions on Deep Synthesis Internet Information Services (2022)
Governs deepfake and synthesis technologies. Requires consent to recreate a real person, labeling, and security assessment for certain services.
Effective: January 10, 2023
Recommendation Algorithm Provisions
Provisions on Algorithmic Recommendation (2021)
Requires recommendation algorithms to be transparent, controllable, and filed with regulators, and gives users the ability to opt out.
Effective: March 1, 2022
Implementation Opinions on the Standardized Application and Innovative Development of Intelligent Agents
CAC, NDRC, and MIIT, jointly issued May 8, 2026
A national framework that regulates agentic AI as its own category. Defines agents by autonomous perception, memory, decision-making, interaction, and execution; calls for clear boundaries between decisions reserved to the user, decisions the user authorizes, and decisions the agent makes on its own, with users entitled to know about and make final decisions on autonomous ones; supports mandatory standards for agents in healthcare, transportation, media, and public safety; and puts agents in sensitive fields and key industries under filing, testing, and recall of problematic products.
Effective: Issued May 8, 2026
Interim Measures for the Administration of AI Anthropomorphic Interactive Services
CAC with four partner agencies, issued April 10, 2026; effective July 15, 2026
The binding companion-AI rule. It is a separate instrument from the agent framework, and the two are often conflated. Providers of human-like interactive AI must run security assessments, must not design for addiction or dependency, must not offer virtual intimate relationships to minors, and must intervene when users show signs of harm. Its July 15 effective date drove immediate feature shutdowns at major Chinese consumer AI products.
Effective: July 15, 2026
Timeline
- March 1, 2022
Recommendation Algorithm Provisions take effect.
- January 10, 2023
Deep Synthesis Provisions take effect.
- August 15, 2023
Interim Measures for Generative AI Services take effect.
- September 1, 2025
AI content-labeling Measures take effect.
- November 25, 2025
The CAC publishes the first batch of enforcement cases under the labeling Measures, with apps ordered to correct or taken down for missing labels.
- January 1, 2026
Cybersecurity Law amendment referencing AI governance takes effect.
- May 8, 2026
Intelligent-agent framework issued jointly by the CAC, NDRC, and MIIT, treating agentic AI as its own regulated category.
- July 15, 2026
Interim Measures for anthropomorphic interactive AI services take binding effect; companion products pull features the same day.
- September 14, 2026
TC260 releases version 3.0 of the AI Safety Governance Framework, a non-binding update to the risk classification and response measures.
- September 15, 2026
The CAC publishes further enforcement cases under the labeling Measures, ordering a generative AI mini program offline.
- September 18, 2026
The CAC opens consultation, to October 17, on a draft State Council regulation on minors online that would bar virtual-companion services for minors across all network services.
- September 30, 2026
A two-month Qinglang campaign targets fake personas on short-video platforms, including AI-synthesized impersonation of public figures; labels such as AI-generated do not excuse platforms from content duties.
- October 23, 2026
The national informatization plan for 2026 to 2030 calls for foundational AI legislation and for assessment and filing of AI agents, on-device models, and anthropomorphic services; the State Council General Office publishes opinions on AI and employment, dated October 16.
- November 2, 2026
The NPC Standing Committee opens a session, to November 6, whose recommended agenda includes the State Council's request to ratify the agreement establishing the World AI Cooperation Organization.
Who it applies to
Providers of internet information services and public-facing generative-AI services in mainland China, including foreign providers reaching Chinese users. Obligations often stack across cybersecurity, data-security, and content rules at once.
Penalties
Enforcement runs through the CAC and sector authorities under the Cybersecurity Law, Data Security Law, and Personal Information Protection Law. Sanctions include service suspension, takedowns, fines, and, after the Cybersecurity Law amendment in force since January 1, 2026, more immediate penalties for serious breaches. Labeling enforcement has run since November 2025; in cases published September 15, 2026 the CAC ordered a generative AI mini program offline for missing explicit and implicit labels and for skipping security assessment.
Recent developments
October 23, 2026
Cyberspace Administration of China
China's informatization plan for 2026 to 2030 calls for foundational AI legislation
The national informatization plan for the 15th five-year period, published by the CAC, calls for advancing foundational legislation on AI alongside the digital economy, telecommunications, and cybercrime, for optimizing the filing system for generative AI, and for assessment and filing of new forms such as AI agents, on-device large models, and anthropomorphic interactive services. The same day the State Council General Office published opinions on AI and employment, dated October 16, that call for employment-risk assessments of AI projects in market-access fields that could cause large job losses and for action against algorithm-driven hiring discrimination.
September 30, 2026
Cyberspace Administration of China
China opens a two-month campaign against fake personas on short-video platforms, AI impersonation included
The Central Cyberspace Affairs Commission Office launched a two-month Qinglang campaign against fake personas on short-video platforms. One of its five targets is impersonation, including AI-synthesized audio and video of public figures used for fake endorsements and marketing. Platforms are told that content labels such as fictional, AI-generated, marketing, or reposted do not excuse them from content-governance duties, and that unlabeled fake-persona videos are to be handled strictly.
September 15, 2026
Cyberspace Administration of China
China orders a generative AI mini program offline in its latest content-labeling enforcement round
In its latest batch of enforcement cases, following rounds published in November 2025 and April 2026, the CAC said a Sichuan company's WeChat mini program offering AI chat and image generation had added no explicit labels, no implicit metadata labels, and no labels on exported content, and had not completed the required security assessment; the program was ordered offline. A Jiangsu company running websites that relayed calls to several large-model APIs was sanctioned for skipping security assessment under the generative AI measures. Three days later the CAC opened consultation, to October 17, on a draft State Council regulation on minors online that would bar virtual-companion services for minors on every network service.
September 1, 2025
China Law Translate
China's AI content-labeling rules take effect with dual visible and embedded marks
The CAC's labeling Measures now require explicit labels users can see and implicit labels inside file metadata on AI-generated content, backed by a mandatory national standard. Major platforms issued compliance statements as the rules went live.
Common questions
Does China have a comprehensive AI law?
Not in one statute. China has an extensive set of binding AI rules without a single comprehensive statute. Instead the Cyberspace Administration of China (CAC) and partner agencies issue targeted rules, each aimed at a specific problem: recommendation algorithms, deep synthesis, generative AI services, and content labeling. Public-facing generative services must pass security assessment and file with regulators, and AI-generated content must carry both visible and embedded labels. A comprehensive national AI law has been on the agenda since 2023 but remains a proposal.
What are the penalties for AI violations in China?
Enforcement runs through the CAC and sector authorities under the Cybersecurity Law, Data Security Law, and Personal Information Protection Law. Sanctions include service suspension, takedowns, fines, and, after the Cybersecurity Law amendment in force since January 1, 2026, more immediate penalties for serious breaches. Labeling enforcement has run since November 2025; in cases published September 15, 2026 the CAC ordered a generative AI mini program offline for missing explicit and implicit labels and for skipping security assessment.
Who does AI regulation in China apply to?
Providers of internet information services and public-facing generative-AI services in mainland China, including foreign providers reaching Chinese users. Obligations often stack across cybersecurity, data-security, and content rules at once.
Cite this page
Artificial Intelligence Regulations, "AI Regulation in China," reviewed October 26, 2026, https://artificialintelligenceregulations.com/jurisdictions/china.html.
Entries state the position as of the review date and link to the primary text for verification.
