# China's Top Spy Chief Flags OpenAI and Anthropic Models as National Security Risks **Chen Yixin**, head of China's **Ministry of State Security (MSS)**, has issued a stark warning regarding the cybersecurity implications of advanced artificial intelligence models. In an article published in the journal of the **Cyberspace Administration of China (CAC)**, Chen highlighted two specific U.S. AI models – **Anthropic's Claude Mythos** and **OpenAI's GPT-5.5-Cyber** – as examples of a "disruptive upgrade" in cyber capabilities. ## The AI Threat Landscape According to Beijing Chen's primary concern revolves around the potential for these AI technologies to be abused by "hostile forces and individuals with ulterior motives." He stated that such abuse could lead to the fabrication of political rumors, the spread of harmful information, and the incitement of confrontational sentiments at low cost and high volume, directly threatening China's "political security, institutional security, and ideological security." He emphasized that cybersecurity is entering a new phase characterized by "vulnerability industrialization, fully automated attack and defense, and AI versus AI." This, he argued, empowers certain entities to rapidly discover vulnerabilities, automate attack paths, and complete complex hacking tasks, thereby lowering technical barriers and costs for cyberattacks against China's critical information infrastructure. ## Echoes of Western Concerns and Internal Disclosures These concerns mirror sentiments expressed by Western governments. The **Five Eyes** intelligence alliance, for instance, warned in June that frontier AI models could reshape offensive and defensive cyber operations within months. While vulnerability reports have surged, a proportional increase in observed cyberattacks has yet to materialize. Notably, Chen's warning came just days after **Anthropic** released a threat report detailing a Chinese-speaking group's use of **Claude** for an "autonomous vulnerability research program." This group, including two university undergraduates, reportedly uncovered several zero-day flaws in a major security product. Earlier reports by **Recorded Future News** also revealed leaked Chinese technical documents that appeared to show a state-backed training platform for rehearsing cyberattacks against critical infrastructure in neighboring countries, potentially involving AI support. ## China's AI Governance and Global Ambitions Beyond cybersecurity, Chen also cited espionage, data leaks, U.S. technology controls, alleged monopolistic practices, algorithmic decision-making in "social governance," and AI's impact on military operations as significant concerns. However, he did not address China's own use of AI in offensive cyber operations, which Beijing consistently denies despite attributions from Western researchers. A day after Chen's article, the **CAC** released an updated AI governance framework during National Cybersecurity Week. This framework focuses on autonomous agents and embodied AI, identifying "loss of control" as a core risk. While currently guidance, China already mandates government security reviews and registration for public-facing AI services. Chen also criticized foreign export controls and "closed-source ecosystems." Chinese labs have increasingly championed open-weight models, a strategy analysts believe aims to spread Chinese technology and influence global technical standards. Chen positioned AI as an "international public good," advocating for its use to help the Global South close the "intelligence gap" and presenting China as an open alternative to what Beijing describes as Western technological monopolies.