China’s spy chief names U.S. AI models as threats to national security

China's top intelligence official has warned that artificial intelligence poses a direct threat to the Chinese Communist Party's authority, in what amounts to Beijing's most sweeping high-level statement on the security risks it associates with the technology.

State Security Minister Chen Yixin, writing in the government-affiliated magazine China Cyberspace on Sunday, called AI "a new arena for strategic rivalry among major powers" and said "hostile forces" were abusing generative AI to "fabricate political rumours and spread harmful information." He warned such actors were launching "public opinion wars and cognitive warfare against China, directly threatening our political security, institutional security and ideological security."scmp+1

Naming American AI Systems

Chen singled out Anthropic's Claude Mythos and OpenAI's ChatGPT-5.5-Cyber by name, citing their advanced hacking capabilities as systems that "pose serious risks to China's critical information infrastructure." He also warned that foreign intelligence agencies were using AI-powered web crawlers, data mining, and profiling tools to collect sensitive state data, trade secrets, and personal information.newindianexpress+2

Rather than echoing Western calls to slow AI development, Chen urged greater financial commitment to Chinese model makers, computing capacity, and data. He wrote that China must "ensure independent control of key core technologies," including the advanced chips vital to training AI models.newindianexpress

Cold War Framing

Chen's article arrived as Beijing's state media mounted a broader counteroffensive against the weekend push by U.S. AI executives — led by Anthropic CEO Dario Amodei — to slow frontier AI development. An editorial in the state-backed Global Times characterized the effort as an "attempt to curb China's AI development through technological barriers and regulatory monopolies" and called it a "silent AI Cold War" that was "hypocritical and short-sighted."cryptopolitan

Chinese analysts quoted by state media went further, with Xiao Qian of Tsinghua University arguing that Anthropic's closed-model approach was "losing ground to Chinese open-source systems on cost, performance, and developer reach." The Global Times counted 12 mentions of China in Amodei's essay and highlighted his own line that his proposed measures "would slow China's progress enough to widen America's lead significantly over the next 3-5 years."cryptopolitan

Ahead of Trump-Xi Talks

The warning sets a confrontational tone ahead of a planned meeting between President Trump and Chinese President Xi Jinping on Sept. 24 in Washington, where AI governance is expected to feature in discussions. Reuters reported that China has been building its own regulatory framework around AI risks, including loss-of-control scenarios, with the Cyberspace Administration of China releasing expanded safety guidelines in September 2025 and joint rules for AI agents issued in May.nytimes+2

Trump, for his part, dismissed the safety warnings. "Whoever wins AI, wins," he told reporters at his golf club in Doonbeg, Ireland, adding that "very negative forces" were raising concerns about things "that won't happen."wionews