A pretraining researcher who has spent the past three years working at two of the world’s most prominent artificial intelligence labs has walked away from his job, delivering a stark public warning about the direction the industry is heading. Jacob Coxon, who most recently worked at Anthropic after an earlier stint at OpenAI, announced his resignation in a lengthy thread on X, accusing both companies of pursuing self-improving AI systems without adequate safeguards.
Coxon’s departure has quickly become one of the most talked-about moments in the ongoing debate over AI safety, largely because of his claim that the danger is not confined to public messaging or activist rhetoric. According to him, the people actually building frontier AI systems privately hold far graver fears than they express to the press.
Researcher Says AI Industry Is “Gambling With Our Lives”
In his resignation statement, Coxon said neither Anthropic nor OpenAI is behaving responsibly as both race toward increasingly capable, self-improving systems. He described the competition among leading labs as a “hubristic gamble” that should not be left to internal corporate decision-making alone.
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 drew a distinction between the two companies. At OpenAI, he suggested, many staff members have not fully absorbed the civilizational stakes of the technology they are building. At Anthropic, by contrast, he said the risks are well understood internally, but the company has convinced itself that it has no choice but to keep pace, reasoning that a less cautious competitor would otherwise take the lead.
“These will soon be superhuman systems that can hack anything, revolutionise any field overnight, and acquire real power and resources,” Coxon warned, adding that progress across these capabilities shows no sign of slowing.
“It Could Kill Us All,” Insiders Privately Believe, Says Coxon
The most striking element of Coxon’s statement was his assertion about what insiders genuinely believe rather than what they say publicly. He argued that many executives and senior researchers at leading AI firms privately fear that the technology could pose an existential threat to humanity before the end of the decade, even as they soften their language for public audiences. “The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote, insisting the claim was “not a marketing stunt.”
The remarks have drawn a response from within Anthropic itself. Evan Hubinger, the company’s Alignment Science Lead, that he and his colleagues do believe AI could pose a lethal risk to humanity, estimating the probability at more than 10 percent within the next decade. Hubinger said Anthropic is “trying its best” on the alignment problem but conceded the company does not yet have a clear plan to ensure safety as systems approach superintelligence, and is not “clearly on track” to solve it. Coxon’s exit adds to a string of recent high-profile departures from AI safety teams, reviving scrutiny over whether the race to build more powerful models is outpacing the industry’s ability to control them.