AI Researcher Resigns From Anthropic Warning Colleagues Fear 'Endgame'
Jacob Coxon's public resignation from Anthropic has renewed scrutiny of internal warnings about self-improving superintelligence, with claims that staff stay out of fear of ceding the field to less scrupulous rivals.
A researcher's public resignation from the American artificial intelligence laboratory Anthropic has intensified debate over the risks of pursuing self-improving superintelligence, after he claimed colleagues share his fears but remain because they are afraid of leaving the field to rivals with fewer scruples.
Jacob Coxon announced his departure in a post on X, arguing that he could no longer work for companies he accused of «gambling with our lives». His intervention has cut through where warnings from more prominent figures have not, according to the Guardian columnist Gaby Hinsliff, because it describes an industry that appears to be begging to be saved from itself.
The most striking element of Coxon's account is his claim about the reasoning of those who stayed. He said colleagues at Anthropic are not staying because they believe he is wrong about the dangers of pursuing self-improving superintelligence — machines not only smarter than humans but capable of building their own more powerful successors and evolving independently of human control — but because they fear ceding the field to people with fewer scruples.
According to Coxon, his former colleagues talk routinely about the «endgame» or «crunch time», meaning that what they do in the next year or two will decide the fate of humanity. Whether that is true or simply self-aggrandising delusion, the picture he paints is of a sector whose own insiders believe the stakes could hardly be higher.
The resignation lands amid a growing volume of warnings about artificial intelligence's existential threat to humanity. Tech whistleblowers have appeared with increasing regularity to warn that their work is probably going to kill us, to the point where the public has become almost blase about the prospect of humanity being wiped out within a decade.
Part of the difficulty for outsiders is that it is never clear whether the tech threat is real or merely a twisted form of hype from an industry that drums up investment by making its products sound more powerful than they really are. Most people settle for trying not to think about it too hard.
Coxon's decision to resign very publicly, rather than quietly move on, has changed that calculation for some observers. His claim that colleagues are staying out of competitive fear rather than scientific disagreement suggests that the people closest to the technology believe the risks are genuine, even if they cannot agree on what to do about them.
The episode also raises questions about how decisions inside leading AI laboratories are made and communicated. If staff believe the next year or two will determine the fate of humanity, the governance of those labs becomes a matter of public consequence rather than private corporate judgment.
Anthropic has not publicly responded to the substance of Coxon's claims. The company has previously positioned itself around safety research, arguing that advanced AI systems require careful development and oversight.
For policymakers and the wider public, the resignation adds to a pattern of insiders warning about risks they say are poorly understood outside a small circle of researchers. The challenge is distinguishing genuine alarm from marketing, and deciding who — if anyone — has the power to pause research that many believe could reshape the world.
Hinsliff's argument is that the warnings are piling up and it is time to listen and take them deadly seriously. The question now is whether a single resignation can shift an industry that has so far treated existential risk as a problem to be managed rather than a reason to stop.
