← All jurisdictions
China
AI-content disclosure obligation
Measures for Labeling AI-Generated Content + mandatory standard GB 45438-2025 (CAC)
Not yet reviewed by an attorney. Primary text last checked September 1, 2026. This summary may be incomplete — see the open items below.
Mark the output as AI-generated. This duty falls on both the AI tool/platform and the publisher. A platform's own automatic AI label does not satisfy this duty on its own.
- Effective
- 2025-09-01
- Applies to
- both
- Extraterritorial
- No
- Penalty
- Not confirmed
Notes
- Real enforcement confirmed: CAC penalized CapCut, 'Cat Box,' and Dreamina AI in April 2026 for label violations.
Open items before this is fully reliable
- This is the fourth and most specific of four related instruments (Deep Synthesis Provisions, eff. 2023-01-10; Generative AI Measures, eff. 2023-08-15; these 2025 Measures + GB 45438-2025) — the earlier two established the general labeling duty this row's 'dual-track' mechanism (visible label + embedded metadata/watermark) actually operationalizes. Not modeled as separate rows.
- Exact technical format of the embedded metadata/watermark not specified.