China's Top Spy Chief Flags US AI Models as Cyber Risk

Chen Yixin, head of China's Ministry of State Security, has publicly named two US AI models, Anthropic's Claude Mythos and OpenAI's GPT-5.5-Cyber, as signs of a 'disruptive upgrade' in cyber capabilities. Writing in a journal published by the Cyberspace Administration of China, Chen said these models could increase the speed and scale of vulnerability discovery and malware development, though he stopped short of claiming either had been used in attacks against China.

Chen described cybersecurity as entering a new era of 'vulnerability industrialization' and 'AI versus AI' conflict, warning that some nations now have the ability to rapidly find flaws and automate complex hacking tasks. This, he said, lowers the technical skill and cost required to launch cyberattacks, posing risks to critical infrastructure.

The warning mirrors concerns raised by Western intelligence agencies. The Five Eyes alliance cautioned in June that frontier AI models could reshape offensive and defensive cyber operations within months. While vulnerability disclosures have reportedly hit record levels since then, a matching rise in actual cyberattacks has not yet been confirmed.