Young and talented, he has now resigned, giving up his brilliant career and leaving a worrying message to the whole world. “Today I resigned from Anthropic. I have spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is behaving responsibly. They are racing headlong into self-improving super intelligence and playing with our lives.” With these words, published this morning on his X account, researcher Jacob Coxon announced his exit from the world of Artificial Intelligence. He leaves the San Francisco laboratories denouncing catastrophic dangers.
Self-improving models
Coxon, a 27-year-old British man with a mathematical background, explained that he is leaving Anthropic because he does not want to participate in the competition to build systems capable of self-improvement. A race that, according to him, could lead to artificial intelligence models running amok. The researcher added that in the sector there is now open talk of “critical period” and “final phase”, expressions that describe the approach to systems capable of exceeding human capabilities.
Coxon’s resignation on ethical grounds comes as Anthropic and other industry players are busy developing ever more powerful models. His publicly motivated farewell is part of the growing debate on security, governance and the need for a coordinated slowdown in the development of artificial intelligence.
“An arrogant bet”
In his posts, Jacob Coxon challenged the logic of accelerated competition: “Accepting this race and entering the ‘final stage’ is an arrogant gamble that should not be made by privately held Slack. Attempting to complete alignment in record time should require extraordinary faith that no better trajectories exist.” This is a direct reference to internal channels of companies where, according to him, strategies that could have global consequences are discussed.
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
The most alarming part of his messages concerns the perception of risk among researchers themselves: “The people developing artificial intelligence sincerely believe that it could kill us all by the end of the decade. This is not a marketing gimmick. In fact, many high-level executives and researchers will try to use more sensible language in their statements to the press, but I hear the same people expressing fear.”
The appeal to researchers
The young researcher had a direct message to colleagues: “If you are a laboratory researcher, I urge you to think about what the next few years will really look like for you. Do you want to start a super-intelligent reinforcement learning experiment without a rigorous understanding of its mind? Should you keep your head down because ‘this is going to happen anyway’ or seize this moment?”