Ex-Anthropic's Jacob Coxon says he's 90% certain PRC has spies in US AI giants

Former Anthropic researcher Jacob Coxon told entrepreneur Patrick Bet-David on the PBD Podcast that he is "90% certain" Chinese intelligence operatives are embedded at two leading U.S. AI companies, naming OpenAI and Anthropic as the firms where he worked on pre-training research. He argued that such infiltration, combined with weak background checks, means Beijing could siphon breakthroughs regardless of whether American developers slow their pace. Coxon did not present evidence or specific names to substantiate the claim, but framed it as a practical counterpoint to proposals that the U.S. can preserve a technological edge simply by accelerating domestic AI development.

Coxon linked the espionage allegation to a broader safety and national-security stance: even a temporary U.S. lead is fragile because rivals could "steal our progress" through hacking or copying, while a deliberate slowdown might also impede China’s ability to catch up. He resigned from Anthropic in September, publicly warning that companies are racing toward self-improving superintelligence and calling the technology a "super weapon." His statements amplify ongoing policy debates over balancing rapid innovation with stronger safeguards and supply-chain or personnel security, adding unproven but urgent espionage concerns to deliberations about AI governance.