What happened
On September 14, 2026, at the opening ceremony of China's 2026 National Cybersecurity Awareness Week, TC260 (the national cybersecurity standardization committee, operating under CAC guidance) published the 'Artificial Intelligence Safety Governance Framework 3.0,' the third annual major revision following v1.0 (Sept 2024) and v2.0 (Sept 2025). The new version updates AI risk classifications and revises the technical-response and comprehensive-governance measures, continuing the framework's 'risk classification, technical response, comprehensive governance' structure, developed with the China Academy of Cyberspace Studies and CAC's Data and Technology Support Center. It implements China's Global AI Governance Initiative.
Why it matters
This is China's principal national AI safety/security governance reference document, directly shaping compliance expectations for AI model developers, service providers and downstream users operating in or serving the Chinese market — including foreign frontier labs whose models or APIs are used or scrutinized in China. Major-version updates historically drive changes in TC260's binding technical standards (e.g., Basic Security Requirements for Generative AI Services) and are watched closely by multinational compliance teams.
Action needed
Map new risk-classification and technical-response provisions in Framework 3.0 against existing AI governance programs (ISO 42001, NIST AI RMF) for organizations with China market exposure; monitor for follow-on binding TC260 standards implementing the framework.