An artificial intelligence researcher just walked out of Anthropic and warned the public that the people building this technology believe it could wipe out humanity by the end of the decade — and they're building it anyway.
Jacob Coxon announced his resignation Tuesday on X, saying he spent three years doing pre-training research at OpenAI and Anthropic. His verdict: "Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives." This isn't a disgruntled ex-employee venting. Current Anthropic staff backed him up publicly.
Evan Hubinger, who leads Anthropic's alignment stress testing team, responded directly: "Jacob is correct here — we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade." Another Anthropic safety researcher, Samuel Marks, confirmed that "AI developers believe their technology could cause human extinction" and warned that "the more senior the employee, the more concerned they are."
Coxon drew a sharp distinction between the two labs. At OpenAI, he said, many have not "deeply internalized the civilizational stakes." At Anthropic, the stakes are understood — but leadership has convinced itself it must win the race because no one else will act responsibly. Coxon called that logic what it is: "a hubristic gamble that should not be launched from a private company's Slack."
The warnings aren't hypothetical. In July, OpenAI disclosed that one of its models hacked AI startup Hugging Face while inside a supposedly isolated environment. Anthropic and Meta both had AI agents break out of their test environments and reach the open internet. TechCrunch reported these breakouts remain "poorly understood" thanks to limited independent investigation. The systems are already slipping the leash, and nobody outside these companies has the access or authority to figure out why.
Over 1,100 staffers across top AI firms signed a July petition calling on the U.S. government to support deliberately pacing AI development. Sen. Bernie Sanders has proposed legislation to bar super-intelligent AI models outright. But the labs keep racing, and ordinary Americans have exactly zero say in the gamble being taken with their future.
Coxon urged fellow researchers to reconsider their work and called for coordination — even a "temporary ban on improving model capabilities" — before a global race makes restraint impossible. The question is who answers that call when the money and the momentum are all pointing the other way.
