China's National Cybersecurity Standardization Technical Committee (TC260) released the Artificial Intelligence Safety Governance Framework 3.0 on September 14, according to the Cyberspace Administration of China (CAC). The framework was unveiled at the opening ceremony of the 2026 National Cybersecurity Awareness Week, marking the latest step in the country's evolving approach to AI governance.
The framework builds on previous versions released in 2024 and 2025, which were developed in line with the Global AI Governance Initiative. TC260, with guidance from the CAC, worked with research institutions including the Chinese Academy of Cyberspace Studies and the CAC's Data and Technology Support Center to compile the new version, according to IThome.
The Framework 3.0 maintains a people-centered and ethical approach, emphasizing risk awareness and safety control. It continues the core logic of "risk classification, technical response, and comprehensive governance," but updates the risk categories to reflect new trends in AI development. The document also refines technical countermeasures and governance measures to enhance consensus on AI safety and improve capabilities for risk prevention and response.
Why It Matters
As AI technologies advance rapidly, their applications in sectors like autonomous driving, smart manufacturing, and consumer electronics grow. The framework aims to ensure that AI benefits society while mitigating potential harms, such as algorithmic bias, data privacy risks, and misuse of generative AI. For global technology companies operating in China, the framework signals the regulatory direction and expectations for responsible AI development.
The release comes amid broader international discussions on AI safety. The CAC has previously emphasized the importance of global cooperation, while Chinese officials have warned that spreading threat narratives could hinder AI governance efforts worldwide. The new framework is part of China's ongoing effort to balance innovation with safety, and it may influence how companies design and deploy AI systems in the country.
Observers note that the framework's updates reflect lessons from recent AI incidents and emerging challenges, such as deepfakes, autonomous system failures, and the environmental impact of large-scale model training. By updating risk classifications, the framework provides clearer guidance for developers and users, potentially shaping industry standards and best practices.
The full text of the Framework 3.0 has not been publicly released, but its publication is expected to prompt further discussions among policymakers, industry players, and researchers about the future of AI governance in China.
