China Measures for Labeling AI-Generated Synthetic Content
Requires AI-generated or synthesised text, images, audio, video and virtual scenes in China to carry labels: visible (explicit) labels where the deep synthesis rules require them, and hidden (implicit) labels in file metadata. Platforms that distribute content must check metadata and add notices. Removing or forging labels is banned.
At a glance
In force
- Jurisdiction
- China
- Type
- Regulation
- Official title
- 人工智能生成合成内容标识办法 (Measures for Labeling Artificial Intelligence-Generated Synthetic Content), CAC, MIIT, MPS and NRTA, Guo Xin Ban Tong Zi [2025] No. 2
What this means for you
If you build AI
- Generation services must write an implicit label into file metadata (content attributes, provider name or code, content ID); digital watermarks are encouraged.
- Where the deep synthesis rules apply, add visible labels to text, audio, images, video and virtual scenes, and keep them in downloaded or exported files.
If you deploy AI
- Platforms that distribute content must check metadata for implicit labels and add a notice around content marked as AI-generated.
- If a user asks for output without a visible label, the user agreement must set their labelling duties and logs must be kept for at least six months.
If you use AI
- When you post AI-generated content on a platform in China, you must declare it and use the platform's labelling tool.
- Maliciously removing, altering, forging or hiding AI labels is prohibited.
Key dates
- Sep 1, 2025PastLabelling measures take effectSource
Status history
In the Drops
No Drop has covered this law yet.