Key facts
- Former Anthropic and OpenAI researcher Jacob Coxon claims Chinese spies are likely present in US AI labs.
- Coxon stated he is 90% certain of Chinese intelligence infiltration at OpenAI and Anthropic.
- He warned that stolen research could neutralize any technological lead held by the US.
- Coxon suggested AI labs have minimal security vetting compared to defense contractors.
- He called for defense-level security protocols and mandatory background checks at AI firms.
Jacob Coxon, a former researcher at both OpenAI and Anthropic, has alleged that Chinese intelligence has infiltrated leading US artificial intelligence laboratories. Speaking on the PBD Podcast and before the New York City Council, Coxon stated he was "90% sure" that spies were present within these organizations.
Coxon warned that insider access could enable China to conduct covert exfiltration and steal AI models, thereby neutralizing any technological lead the United States might build. He drew a comparison to defense contractors like SpaceXAI, which undergo stringent background checks and hiring restrictions for foreign nationals, arguing that Silicon Valley AI labs currently operate with minimal security vetting. Coxon advocates for implementing defense-grade counter-intelligence protocols, mandatory background checks, and potentially restrictive hiring rules.
The allegations surface amid intense competition between the US and China in the AI sector. Both nations are rapidly developing and releasing new AI models. Coxon's concerns highlight the difficulty of protecting valuable research as AI models become more sophisticated, especially if sensitive advancements can be easily replicated.
