This living tracker separates effective national AI-related rules from proposals and policy programmes. As of 31 August 2026, China uses a layered framework for algorithmic recommendations, deep synthesis, public-facing generative AI, AI-generated-content labels and anthropomorphic emotional-interaction services.
Use the tracker to identify an instrument and effective date, then complete a product-specific applicability analysis. Filing, safety assessment, labeling, privacy, IP and sector duties depend on the actual service, operator, users and functions; a product described commercially as “AI” does not automatically fall within every row.
Effective national framework
| Instrument | Effective | Operational significance |
|---|---|---|
| Algorithmic Recommendation Provisions | 1 March 2022 | Governance, user controls, transparency, filing and security-assessment framework for covered internet information services |
| Deep Synthesis Provisions | 10 January 2023 | Identity, consent, training-data, technical review, safety assessment, logs and synthetic-content labels |
| Interim Generative AI Measures | 15 August 2023 | Public-facing generative-AI provider duties, training inputs, output governance, user protection, assessment and filing |
| AI-Generated and Synthetic Content Labeling Measures and GB 45438-2025 | 1 September 2025 | Explicit and metadata labels, distribution-platform handling, user declarations, export controls and anti-tampering |
| Anthropomorphic AI Interaction Measures | 15 July 2026 | Emotional-interaction safety, dependence controls, minors, interaction data, notices, exit, assessment and filing |
Filing and registration status
CAC's public announcement is a living source for filed generative-AI services and registered applications or functions that directly call filed model capabilities. Its attachments were updated through February 2026 when this tracker was checked. A team should verify the current attachment and relevant provincial cyberspace administration record before launch or diligence.
Record the model name, provider, filing or registration number, operator entity, application or feature, publication location and date checked. A vendor screenshot is not a substitute for the official record, and an entry does not remove other service, data, IP, consumer or sector obligations.
2025 labeling milestone
From 1 September 2025, covered providers and dissemination platforms must implement the labeling measures together with mandatory GB 45438-2025. The key control change is that labeling is not only a visible watermark: it combines medium-specific explicit notices with provenance information in file metadata and assigns duties to propagation platforms and users.
Release evidence should include screenshots or recordings of explicit notices, metadata samples for every supported format, export and API tests, platform ingestion tests, user-declaration controls, service terms, six-month log handling for the limited no-explicit-label route, and anti-removal monitoring.
2026 anthropomorphic-interaction milestone
From 15 July 2026, providers of continuing emotional-interaction services that simulate a natural person need a dedicated control set. Applicability turns on the emotional and continuing nature of the interaction. Ordinary customer service, knowledge Q&A, work assistants, education and scientific research are expressly excluded where they do not involve continuing emotional interaction.
Change review should therefore examine persistent persona, memory, relationship framing, emotional dependence, rewards, re-engagement, virtual intimacy, minor access and continuous-use duration. Adding these functions can change the regulatory analysis even if the base model and company remain the same.
Legislative and enforcement watchlist
- Comprehensive AI legislation: the State Council's 2026 legislative work plan calls for accelerating work, but this tracker does not treat a consolidated AI act as enacted.
- Filed-service lists: monitor CAC and provincial updates and reconcile product displays to the current model and operator.
- Label enforcement: retain technical evidence across all supported formats and third-party distribution paths.
- Product changes: reassess filing, safety assessment and labeling when a model, operator, user group, feature or distribution mode materially changes.
- Sector rules: health, finance, vehicles, education, advertising and employment require separate monitoring.
Internal tracker workflow
- Capture the official instrument, issuer, publication date, effective date and official URL.
- Classify it as effective, transitional, draft, standard, guidance, filing notice or enforcement action.
- Map affected products, entities, roles and owners.
- Translate each requirement into a control, evidence item and deadline.
- Record counsel confirmation where interpretation or filing interaction is required.
- Close the change only after product evidence and public disclosures are tested.
For the operating control map, use China's AI Rules: A Practical Business Guide.
Official sources
- Interim Measures for Generative Artificial Intelligence Services
- Provisions on Algorithmic Recommendations in Internet Information Services
- Provisions on Deep Synthesis in Internet Information Services
- Measures for Labeling AI-Generated and Synthetic Content
- GB 45438-2025 labeling standard
- CAC generative-AI filing and registration announcement
- Interim Measures for Anthropomorphic AI Interaction Services
General information only, not legal advice. Official-source currency checked on 31 August 2026. The existing legal-review attribution is preserved; this editorial update does not represent a new legal review.
