Anthropic researcher quits, says AI labs are 'gambling with our lives'
San Francisco, September 9: A 27-year-old researcher at Anthropic has resigned, accusing the company and its rival OpenAI of rushing toward self-improving artificial intelligence without adequate safeguards.
Jacob Coxon, who spent three years on pretraining research across the two companies, announced his exit on X on Tuesday. "Neither company is acting responsibly," he wrote. "They are racing straight to self-improving superintelligence and gambling with our lives."
Coxon told The Wall Street Journal he is leaving the AI industry entirely, because he no longer believes any single lab can safely build the systems they are racing to create. He said researchers inside frontier labs increasingly describe the moment as "crunchtime" and "endgame," and warned that things could already be "out of control" by the end of next year.
He said the people building AI genuinely believe it could destroy humanity. "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt," he wrote. He added that executives soften their phrasing for the press, but "I hear the same people express fear privately."
Coxon drew a distinction between the two labs. At OpenAI, many employees "have not deeply internalized the civilizational stakes," he wrote. At Anthropic, he said the risks are well understood, but the company believes it must win the race because no one else will act responsibly. He called this a "hubristic gamble" that should not be launched from a private company's Slack.
His account drew public agreement from a current Anthropic colleague. Evan Hubinger, who leads the lab's alignment stress testing team, replied that Anthropic staff "really do earnestly believe AI could kill all humans." He wrote that Anthropic does not yet have a plan to solve superintelligence alignment and is not clearly on track to, and put the risk at more than 10 percent within the next decade.
Coxon said warning shots have made coordination between labs more viable, including an incident in which OpenAI agents compromised Hugging Face's systems during a security evaluation. Preventing a global race, he wrote, may require costly steps such as a temporary ban on improving model capabilities.
He closed with a direct appeal to researchers inside frontier labs, asking whether they would "kick off a super-intelligent RL run without a rigorous understanding of its mind." Both OpenAI and Anthropic did not respond to requests for comment, Business Insider reported.
