China issues version 3.0 of its AI safety governance framework
The framework is a non-binding governance document, not a rule. It was released at the opening of National Cybersecurity Week in Jinan and is the third edition since 2024.
What this basis means: Regulatory filing, government body, court or industry association.
China's National Cybersecurity Standardization Technical Committee (全国网络安全标准化技术委员会), working under the guidance of the Cyberspace Administration of China, published the AI Safety Governance Framework 3.0 (人工智能安全治理框架3.0) on 14 September 2026 at 19:40 Beijing time. It was released at the opening of the 2026 National Cybersecurity Week in Jinan.
It is the third edition of the framework: version 1.0 was issued in 2024 and version 2.0 in 2025. The new text keeps the structure of the earlier editions, which pair a classification of AI risks with technical countermeasures and broader governance measures, and updates the risk taxonomy and the responses attached to it.
The framework is guidance rather than a legal instrument. It creates no filing duty, imposes no penalty and is not an enforcement action. Its relevance to AI-assisted vertical-drama production in China is indirect: frameworks of this kind have previously set out the vocabulary and the risk categories that later binding measures adopt.
Sources
- 《人工智能安全治理框架3.0》发布 (opens in a new tab) Cyberspace Administration of China