China Releases AI Safety Evaluation Framework for Industrial Sector
On October 15, 2026, the 2026 Artificial Intelligence Industry Conference opened at the Yellow River International Convention and Exhibition Center in Jinan, where the "Artificial Intelligence Safety Evaluation Framework for the Industrial and Information Technology Sector (2026 Edition)" was officially released.
The framework was developed by the China Artificial Intelligence Industry Alliance together with relevant organizations, under the guidance and support of the Cybersecurity Administration of the Ministry of Industry and Information Technology. As AI rapidly integrates into industrial and information technology fields, emerging applications such as intelligent agents and embodied intelligence have raised new requirements for security assurance.
Based on the actual development of AI in the sector, the framework establishes a layered and comprehensive safety evaluation system covering general security, application security, and business security, with corresponding evaluation methods and implementation requirements. General security covers data security, model security, and infrastructure security. Application security focuses on intelligent agent security and embodied intelligence security, with requirements for permission security, behavior security, memory security, and tool-call security. Business security covers industrial scenarios and telecom scenarios, spanning the full lifecycle from R&D and design to pilot testing, manufacturing, marketing, and operations management in industry, and covering network planning, operations, maintenance, cloud resource scheduling, and cloud security defense in telecom.
Units participating in the release ceremony included the China Artificial Intelligence Industry Alliance, the MIIT AI Standardization Technical Committee, and the Internet Society of China, along with companies such as China Telecom, China Unicom, Volcano Engine, Baidu, Ant Group, MiniMax, StepFun, 01.AI, Geely, 360, and Angs.
