Jacob Coxon claims Chinese espionage threatens US AI progress

Former Anthropic researcher Jacob Coxon has claimed that slowing the pace of artificial intelligence development in the United States may be ineffective due to potential espionage. Speaking on a podcast hosted by Patrick Bet-David, Coxon stated he is “90%” certain that Chinese spies have infiltrated major AI labs, including OpenAI and Anthropic, to steal progress and close the technological gap.

Coxon argued that whether the U.S. continues rapid development or slows down, China will eventually gain access to the technology. He also cited the use of distillation attacks—where advanced models are used to train less advanced ones—as a method by which Chinese labs, such as Alibaba and DeepSeek, are advancing their capabilities. Anthropic has previously accused several Chinese firms of launching millions of such attacks.

Anthropic has denied the claims of infiltration, stating there is no evidence to support them and noting that the company maintains rigorous personnel vetting and access controls. Coxon did not provide specific names of individuals involved in alleged espionage, stating he was “not trying to, like, start a witch hunt.”