China publishes its third AI safety framework

China published AI Safety Governance Framework 3.0 on September 14, its third version after frameworks 1.0 and 2.0, updating risk categories, technical responses, and governance measures under national cybersecurity standards guidance. A companion official account says 988 generative-AI services had completed filing and 598 AI applications or functions had registered by June 30, putting the revision against a live deployment base. Builders shipping AI into China should treat safety policy as a versioned release artifact—map authority, monitoring, and incident response to the latest framework, then compare it with Beijing’s broader AI blueprint before assuming one compliance memo will age well.