China spy chief points at US AI models in cyber threat warning
China's top intelligence official, Chen Yixin, warned that U.S. AI models, specifically Anthropic's Claude Mythos and OpenAI's GPT-5.5-Cyber, pose a significant cybersecurity risk to China, citing concerns about vulnerability industrialization and AI-powered attacks. This warning comes amidst growing anxieties about AI’s impact on cyber operations globally and China’s efforts to promote open-source AI technology as an alternative to Western dominance. The Chinese government is already implementing AI governance frameworks and reviewing AI services to mitigate potential risks.
China’s top intelligence official, Chen Yixin, has issued a stark warning regarding the cybersecurity risks posed by U.S. artificial intelligence models. In an article published in the journal of the Cyberspace Administration of China, Chen identified Anthropic’s Claude Mythos and OpenAI’s GPT-5.5-Cyber as key concerns, arguing that these models represent a ‘disruptive upgrade’ in cyber capabilities. He described a shift towards ‘vulnerability industrialization,’ where AI can rapidly discover and exploit weaknesses in security products, and automate complex hacking tasks, significantly lowering the barriers to launching cyberattacks.
Chen highlighted concerns about the potential for AI to be used to generate political rumors, spread harmful information, and incite confrontational sentiments – a tactic he attributed to ‘hostile forces and individuals with ulterior motives.’ He emphasized that China faces increasing risks to its critical information infrastructure due to these advancements.
Western governments, including the Five Eyes intelligence alliance, have expressed similar concerns about AI’s impact on cyber operations, noting a surge in vulnerability reports since the introduction of advanced AI models. Despite this increase in vulnerability discovery, a corresponding rise in actual cyberattacks hasn't yet been observed.
The warning comes as China is actively developing and promoting open-source AI models, in part as a strategy to counter perceived Western technological dominance and influence global technical standards. China already requires public-facing AI services to undergo government security reviews and register before launch.
Anthropic recently published a threat report detailing a Chinese-speaking group utilizing Claude to autonomously research vulnerabilities in a major security product, with operators identified as university students in Hunan. The report also documented similar misuse by Russia-linked actors and other groups. The Cyberspace Administration of China released a new AI governance framework during National Cybersecurity Week, focusing on autonomous agents and ‘loss of control’ as a core risk. This framework is guidance, not law, and reflects China’s broader efforts to manage the potential risks associated with AI development and deployment.