Jacob Coxon, a 27 year old researcher who worked on AI pretraining at both OpenAI and Anthropic, resigned from Anthropic on September 8 and left the AI industry entirely, warning that leading AI labs are racing toward self-improving superintelligence without the safety measures to control it.
In a resignation post on X, Coxon wrote that he had spent three years doing pretraining research at both companies and that "neither company is acting responsibly," adding that they are "racing straight to self-improving superintelligence and gambling with our lives." The post reportedly drew more than 90 million views within 24 hours.
What He Told CNN
Coxon appeared on CNN with Anderson Cooper the day after his post went viral. He said today's AI systems do not pose an extinction risk on their own, and that his concern is specifically about the pace at which they are improving, not their current capability. "The people building AI earnestly believe that it could kill us all by the end of the decade," he told CNN.
Coxon told CNN he had originally planned only to quit quietly, but decided to make his concerns public because he wanted his resignation to serve some purpose. He said AI companies are "begging to be regulated" when asked whether their public calls for oversight amounted to more than lip service.
A Current Anthropic Researcher Agreed
Coxon's post drew a notable response from inside Anthropic. Evan Hubinger, described as the company's head of alignment stress testing, wrote on X that Coxon was correct, adding, "We really do earnestly believe AI could kill all humans!" Hubinger put his own estimate of that risk within the next decade at more than 10 percent, while also stressing that present day models do not carry that risk.
Coxon had moved to Anthropic in July 2026, partly drawn by its reputation for taking AI safety more seriously than OpenAI, where he said the internal culture made candid discussion of these risks more difficult. He resigned roughly four months later.
Why This Matters Beyond One Resignation
Anthropic's own chief executive, Dario Amodei, has previously acknowledged similar risks. In comments to the New York Times earlier this year, Amodei said that AI development is "a complex engineering problem" and that he expects something will eventually go wrong with an AI system, hopefully not his own.
Despite these warnings from within the industry, binding regulation remains limited. Efforts to introduce stronger AI oversight in the United States have faced resistance, with critics of regulation often framing it as a disadvantage against China's less constrained AI development.
