China’s AI-generated synthetic content labeling measures came into effect on 1 September 2025. They distinguish visible disclosure from machine-readable metadata and allocate duties across applicable service and dissemination roles.
Two different label types
The CAC regulation identifies explicit labels that users can perceive—text, visual, audio or interface indicators—and implicit identifiers carried in file metadata. The forms depend on the content type, such as text, images, audio, video or virtual scenes. Invisible metadata is not a substitute for every required user-facing disclosure.
Production and distribution responsibilities
Depending on the regulated situation, applicable generative-service providers must incorporate specified indications and metadata, while content distribution platforms have related checking and presentation duties. The measures address user declaration of generated content and prohibit malicious removal or alteration of required labels. Do not generalize these duties to every private image edit everywhere in the world.
Export considerations
A producer should test whether exported images or videos retain the required information after editing, conversion and upload to a platform. Ask what evidence is preserved for creators and reviewers. The regulation contains limited situations addressing requests for exported content without an explicit label under conditions and duties; do not interpret this as a blanket permission to omit public disclosure.
Relationship to provenance standards
C2PA credentials can encode signed media history, but a cryptographic provenance manifest and a jurisdiction’s legal disclosure format are not automatically interchangeable. Map the precise form and technical standard required for the affected service. A media badge that looks credible to a consumer may still omit mandatory file metadata.
Example: online video publisher
A publisher distributing synthetic narration or altered scenes to Chinese users should examine whether the service and content fall within the measures, what disclosures should appear to viewers, what metadata must be preserved, and which platform duties apply. Obtain local compliance review before removing labels or reusing a creator’s likeness.
Regulatory date
The CAC published the measures on 14 March 2025; Article 14 states they take effect on 1 September 2025. They should not be described as law newly created in 2026.
Dated first-party verification and limits
Article 14 of the CAC measures sets 1 September 2025 as the effective date. The obligations turn on the regulated service and dissemination role; the regulation should not be simplified into an assertion that every privately generated image everywhere must carry the same visible mark.
Sources & Further Reading
- Measures for labeling AI-generated synthetic content — CAC China.
- Interim Measures for Generative AI Services — CAC China.
- C2PA specifications — C2PA.
- Measures on labeling AI-generated synthetic content (official Chinese text) — Cyberspace Administration of China.
Guidance varies by jurisdiction and application. This is general education, not legal, medical or other professional advice.