At an AI Security Governance Forum held during China's 2026 Cybersecurity Week in Jinan, Shandong Province, on September 15, Wang Jingtao, deputy director of the Cyberspace Administration of China, called on regulators to practice what he termed "agile governance" and "inclusive prudence" toward artificial intelligence, according to a report by legal and regulatory news service MLex. Wang said authorities should use targeted, flexible tools that contain risks without holding back technological and industrial development. He proposed exploring regulatory sandbox mechanisms that would let companies test new AI systems under controlled conditions, embedding safety standards upstream in how foundation models are developed and trained, and setting clear behavioral boundaries and restricted permissions for autonomous AI agents. Wang also called for closer coordination between AI developers and cybersecurity firms. Separately, the China Electronics Standardization Institute, part of the Ministry of Industry and Information Technology, released a set of AI application security guidelines, made up of a general cross-industry document plus guidance tailored to the education, healthcare, broadcasting, television and online audiovisual sectors. The forum followed China's release of an updated AI Security Governance Framework the day before, on September 14, now in its third version, which places new emphasis on risks from autonomous AI agents. No specific compliance deadlines were announced for the new sector guidelines.
China's Cyberspace Regulator Calls for "Agile Governance" of AI Agents, Issues Sector Security Guidelines
A senior Cyberspace Administration of China official told a Cybersecurity Week forum in Jinan that AI regulators should use flexible "agile governance" tools, including sandbox testing and permission limits on autonomous AI agents, as a standards body issued new sector-specific AI security guidelines, a day after Beijing updated its national AI Security Governance Framework.
Source: MLex, September 16, 2026
Comments
0 commentsNo comments yet — be the first.