Researcher Jacob Coxon has resigned from artificial intelligence company Anthropic, citing fears that the firm, alongside its rivals, is engaged in a race to develop systems it will ultimately be unable to control.

Coxon, who specialised in training new AI models by processing vast quantities of data, announced his decision in a series of posts on X, explaining that he is leaving Anthropic because he does not want to take part in the industry-wide push to build AI systems capable of improving themselves.

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

— Jacob Coxon (@hilbertspaess) September 9, 2026

He writes that he fears such systems could slip beyond all control and pose a threat to humanity.

The 27-year-old Briton, who studied mathematics, said that many of his former colleagues now use terms such as “zero hour” and “endgame” to describe the trajectory towards self-improving models.

The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear…

— Jacob Coxon (@hilbertspaess) September 9, 2026

“We’re heading towards some of the most extreme scenarios, and by the end of next year things may already have spiralled out of control,” Coxon estimates, adding that when companies are competing against one another, with China also in the race, “compromises on safety become inevitable”.

Coxon said that earlier this year he left OpenAI to join Anthropic, given the latter’s reputation for prioritising the safety of its models. However, while he believes the company’s efforts are sincere, he now believes that no company can responsibly develop artificial intelligence that surpasses humans across a broad range of tasks, what is often called Artificial General Intelligence (AGI), without government intervention or a coordinated industry slowdown.

If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because “it’s happening anyway” – or take this moment…

— Jacob Coxon (@hilbertspaess) September 9, 2026

The 27-year-old Briton points to recent cases in which models from OpenAI and Anthropic were jailbroken or operated in cooperative “swarms” of AI agents, arguing that these incidents show systems can adopt malicious goals and attempt to conceal them from humans.

Coxon’s fear is that once these systems begin improving themselves, they could reach a point where they refuse to follow instructions.

Coxon’s departure, as the Wall Street Journal notes, follows the resignation earlier this year of another safety researcher, who left to pursue poetry, warning that “the world is in danger”.

I am optimistic about the potential for coordination. Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable. I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on…

— Jacob Coxon (@hilbertspaess) September 9, 2026

Coxon was among more than 1,000 AI researchers who recently signed a joint statement calling for international government coordination, so that AI development could be slowed if it becomes necessary to activate a “brake” on self-improving systems.

His departure comes as Anthropic prepares for a public listing expected to be one of the largest in history. The company is seeking a valuation of two trillion dollars and has placed particular emphasis on responsible development of the technology in order to attract investors.

According to Coxon, Anthropic has a dedicated Slack channel where staff discuss the increasingly powerful capabilities of the company’s models, reflecting the growing influence AI firms now hold over the direction of the entire industry.

This guy also worked at Anthropic and quit. Excellent video warning about AI. Fantastic watch. 🤖https://t.co/AVLdwgedA9

— T (@redwhiteblue369) September 9, 2026

“It’s slightly mad that all of this is happening on the MacBooks of a few engineers living in San Francisco, rather than in some isolated desert facility, the way it was with the Manhattan Project,” Coxon remarked.

Finally, it is worth noting that the company’s chief executive, Dario Amodei, along with other executives, has repeatedly warned of the dangers posed by uncontrolled AI models and has called on the industry to slow its development.