China opened comment on six draft guidance documents for automotive AI

The intelligent connected vehicle subcommittee of China’s National Technical Committee of Auto Standardization, SAC/TC114/SC34, published six draft national standardisation guidance documents on automotive artificial intelligence for public comment on 10 November, according to its notice. Comments close on 10 December, according to the committee’s comment list. The drafts cover cybersecurity requirements and test methods, data security requirements and test methods, a general development guide for end-to-end models, a test guide for driving automation systems built on an end-to-end architecture, technical requirements and test methods for driving automation systems, and general requirements for risk assessment and governance. They are drafted as guidance technical documents (GB/Z), not as mandatory national standards.

The cybersecurity draft was proposed by the Ministry of Industry and Information Technology. Its requirements fall into four parts: AI process assurance, the AI system itself, cockpit AI applications, and driving automation applications. They include protection against model theft and model inversion attacks and limits on the resources an AI agent can access, the services it can call and the operations it can perform; detection of poisoned training data is recommended rather than required. The explanatory note says the drafting plan was issued on 24 April 2026, and that current automotive cybersecurity standards centre on vehicle networks, electrical and electronic systems and software security. Drafting units named in the cybersecurity draft’s foreword include CATARC, Yinwang, Changan, Li Auto, BYD, Xiaomi, Horizon Robotics and Hesai.

A day earlier, the same subcommittee opened comment until 9 December on Part 4 of its driving automation simulation engineering series, which sets implementation methods and requirements for simulation tests.