Ex-Anthropic Researcher Claims Chinese Spies Infiltrated US AI Giants

Overview

A former researcher at Anthropic has raised alarming claims about the existence of Chinese espionage within America’s leading artificial intelligence laboratories. Jacob Coxon, who departed the company in September, stated he is 90% certain that Chinese spies have infiltrated both Anthropic and OpenAI.

His allegations were made during an appearance on the PBD Podcast, where he argued that the possibility of such infiltration fundamentally undermines the strategy of maintaining a competitive edge through rapid domestic development. Coxon warned that Beijing possesses the capability to steal technological breakthroughs at any moment, regardless of how fast U.S. developers attempt to progress.

The comments add a significant layer to the ongoing national security debate surrounding AI. They suggest that even if American companies achieve a temporary lead, that advantage may be fleeting if sensitive research can be exfiltrated by foreign intelligence operatives.

The Allegations of Corporate Espionage

Coxon’s claims center on the vulnerabilities present within the rigorous hiring processes of top-tier AI firms. During his interview with entrepreneur Patrick Bet-David, he emphasized that current background checks are insufficient for the level of secrecy required by such advanced technology.

"I think one really important thing people don't realize is that China already has spies at OpenAI and Anthropic," Coxon said. "It's almost certain that they have spies already there … There's no serious background check for this sort of stuff yet."

He asserted that Chinese intelligence services have the ability to hack into systems and steal progress at any time. According to Coxon, this capability means that pulling ahead in development does not guarantee a long-term strategic lead. If the United States accelerates its work, China could simply intervene, steal the advancements, and catch up rapidly.

However, Coxon also noted a counterintuitive dynamic. He argued that if the United States were to slow down its AI development, it would likely slow China down as well. This is because, in his view, a significant portion of China’s current progress is derived from copying or stealing U.S. work, even if not all of it.

Context of Resignation and Safety Concerns

These espionage allegations follow closely on the heels of Coxon’s highly publicized resignation from Anthropic. In his departure announcement on X, he cited deep concerns about the direction of AI development and the lack of responsibility shown by major companies.

Coxon stated that he spent the last three years conducting pre-training research at both OpenAI and Anthropic. He concluded that neither organization is acting responsibly, describing their trajectory as a race toward self-improving superintelligence.

"I resigned from Anthropic today," Coxon wrote. "Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."

His recent comments expand on these safety fears by framing AI as a critical national security threat rather than merely a commercial competition. He described the technology as a "super weapon," suggesting that the public and policymakers must soon realize the stakes extend far beyond ordinary business metrics.

Impact on Policy and Industry Debate

Coxon’s statements arrive at a pivotal moment for U.S. policymakers. There is an active debate regarding whether the United States should accelerate AI development to maintain dominance over China or introduce stronger safeguards to mitigate existential risks.

The argument for acceleration is often based on the premise that a significant technological lead provides a buffer against potential threats. However, Coxon’s claims challenge this assumption. If proprietary research can be stolen with relative ease, the value of that lead may be significantly diminished.

This perspective complicates the strategies proposed by leaders who argue that slowing down development could also slow down adversarial nations. Coxon’s insider view suggests that the relationship between U.S. progress and Chinese advancement is more fragile and interconnected than previously understood.

Lack of Evidence and Broader Sentiment

It is important to note that Coxon did not provide specific evidence or name individual spies to substantiate his allegations. His comments remain assertions rather than proven facts. Nevertheless, they resonate with a growing doomsday narrative surrounding the national security implications of advanced AI.

There is increasing competition between the United States and China, and many experts share concerns about the vulnerability of sensitive intellectual property. The notion that AI could be weaponized or stolen is not unique to Coxon, but his position as a former insider lends weight to the discussion.

The conversation also continues regarding whether companies developing increasingly powerful systems can innovate at a responsible pace. Anthropic CEO Dario Amodei has previously discussed the potential for AI to achieve breakthroughs as significant as curing cancer, highlighting the dual nature of the technology as both a beneficial tool and a potential risk.

As policymakers grapple with these issues, Coxon’s warnings serve as a stark reminder of the intersecting dangers of rapid technological advancement and geopolitical rivalry. The question of how to protect America’s AI leadership while managing the risks of uncontrolled development remains one of the most pressing challenges of the decade.