China

• 2026-10-09:The Opinion on Developing New Quality Productive Forcesorders full implementation of the "AI+" action. It is issued policy, not law. The cited Xinhua summary does not support the "safe, reliable, controllable" wording sometimes attributed to it, so that wording isnot verified; the full gov.cn text was not fetched.
• 2026-09-14:TC260 released Framework 3.0(voluntary) and the CAC published its generative AI filing list for July to August 2026 (124 new filings and 133 new registrations).
• 2026-09-11:MIIT issued the "AI+ Software" special action plan(mirror source only).

What to watch:whether a comprehensive AI Law moves from the preparatory stage onto the formal legislative agenda.

China has no single AI statute. Governance comes from a layered set of statutes, administrative measures and policy documents:

• Amended Cybersecurity Law(passed 2025-10-28; in force 2026-01-01). It contains the first statutory-level AI provisions. An earlier wiki entry cited Article 20; that article number is unverified.
• Interim Measures for Anthropomorphic Interactive AI Services(Order No. 21, issued jointly by five regulators; adopted 2026-02-02; in force 2026-07-15). These rules cover companion-style AI that simulates a natural person's personality, and address minors, anti-addiction design and security assessment.
• Generative AI service filing regime.Providers must file with the CAC.
• TC260 AI Safety Governance Framework 3.0(2026-09-14). Voluntary.
• MIIT "AI+ Software" special action plan(2026-09-11), with targets to 2028. The official text was not fetched; only a mirror was read.
• Opinion on Developing New Quality Productive Forces(2026-10-09, CCP Central Committee and State Council). An issued policy document, not law.

The measures apply mainly to providers of generative and interactive AI services offered to the public in China. The Anthropomorphic Interactive AI Measures cover continuous emotional-interaction services that simulate a natural person's personality or style (companion AI), and exclude non-emotional customer-service, question-answering and work-assistant tools.

A comprehensive AI Law appears only as a preparatory item in the State Council's 2026 work plan and in National People's Congress review lists. These are secondary-source reports and official URLs were not captured, so the point needs confirmation.

Generative AI services must be filed with the CAC. The register stood at 1,112 filed services and 731 registered applications as of 2026-08-31, and the CAC published a further filing list on 2026-09-14.

• Companion AI (Order No. 21).Virtual intimate-relationship services are banned for minors, and a minor mode with guardian controls is required. Providers must detect and intervene in extreme emotional states, including by contacting guardians. Design aimed at replacing social interaction, controlling users' psychology or inducing dependence is prohibited, and use reminders are required every two hours. A security assessment is due to the provincial CAC at launch or major change, or where a service has 1 million or more registered users or 100,000 or more monthly active users. Algorithm filing with annual verification applies, and sensitive personal interaction data may not be used for training without separate consent.
• Penalties.Fines of RMB 10,000 to 100,000, rising to RMB 100,000 to 200,000 where life or health is harmed.
• Campaigns.Phase 2 of the "Clean & Clear" AI-application campaign(content-side) began 2026-09-02.

Research assistance, not legal advice; verify before reliance.