At a Sept 1 press briefing for the 2026 National Cybersecurity Awareness Week,
the Cyberspace Administration of China (CAC) identified five principal AI
security risks: 1) innate technical fragility undermining output stability and
reliability; 2) model capability leaps that can upend traditional security
paradigms and produce extreme loss-of-control risks; 3) rapid evolution of
application forms, accelerating iteration of security-risk types; 4) heightened
misuse, abuse and malicious-use risks that threaten social order and ethics
boundaries; 5) risk of technological hegemony in the global AI arena.
