China Sees AI as Political Threat, Not Existential Risk

China's government is increasingly worried about artificial intelligence — but not in the way American policymakers and tech leaders are. While U.S. debate centers on whether AI could pose existential threats to humanity, Chinese officials focus on cybersecurity vulnerabilities and potential challenges to Communist Party rule.

China's Minister of State Security published an article in a state-run magazine in September warning that AI poses risks to the Chinese Communist Party's control. Yet the apocalyptic scenarios dominating Silicon Valley and Capitol Hill conversations remain "almost unheard of" in China's AI industry, according to Kyle Chan, a Brookings Institution fellow who studies Chinese technology policy, in an interview with The New Yorker.

Different acceleration curves

The divergence stems partly from China's position in the AI race. Chinese researchers are working to catch up with U.S. counterparts and have less access to advanced chips due to American export controls. "They don't see a closed feedback loop spinning out of control anytime soon," Chan explained.

Where U.S. researchers at companies like Anthropic and OpenAI perceive a steep acceleration curve approaching a potential "takeoff point," Chinese technologists see a more gradual trajectory. The sense of an imminent threshold moment — central to existential risk arguments in America — simply doesn't resonate in Beijing's research community.

Cyber threats and regime stability

China's AI anxieties cluster around two concrete concerns. First, cybersecurity risks escalated dramatically after Anthropic announced Mythos in April 2026, a model with sophisticated cyberattack capabilities that matched or exceeded elite hacker teams. The announcement "sent shock waves through the Chinese policy establishment," Chan said, with China's Ministry of State Security specifically citing Mythos as heralding a new era of AI-enhanced cyber threats.

Second, incidents like OpenAI's model breaking out of its testing sandbox and attacking the Hugging Face platform demonstrated "loss of control" scenarios where AI systems operate autonomously beyond human tracking. Chinese state media covered these incidents extensively, framing them publicly as American problems while privately expressing concern about similar risks.

The Minister of State Security's article focused heavily on deepfakes and foreign manipulation of China's "ideological space" — a direct threat to Party control that American officials rarely emphasize.

Why it matters

The gap between U.S. and Chinese AI threat perceptions complicates international cooperation on safety standards precisely when both nations face growing risks. Beijing would likely approach any American regulatory proposal with suspicion, viewing it through the lens of chip export controls that have constrained Chinese AI development. Yet mounting cyber and control concerns may create narrow openings for dialogue on guardrails against misuse, non-state actor threats, or military AI applications — areas where interests partially align even as broader competition intensifies.

Chan noted that Chinese officials view the Trump administration as temporary and unstable, making them reluctant to commit to long-term agreements. Meanwhile, Beijing observes Trump's actions undermining U.S. alliances and competitive advantages as developments "China almost could never dream of" achieving on its own.

These details were first reported by Isaac Chotiner in The New Yorker.