Ex-Anthropic Researcher Jacob Coxon Says AI Insiders Fear Human Extinction Within a Decade Jacob Coxon, who spent three years inside OpenAI and Anthropic, resigned with a warning that many AI workers privately believe advanced systems could end human life within about ten years. Anthropic pushed back without disputing the core fear. A researcher who says he spent the past three years working inside both OpenAI and Anthropic has resigned from Anthropic and gone public with an unusually direct warning: a large share of the people building advanced artificial intelligence genuinely believe their own systems could threaten human survival within roughly ten years. Jacob Coxon announced his departure in a series of social media posts that drew millions of views and set off one of the sharpest public arguments yet about the internal culture of the companies racing to build superintelligence. "I believe there is a strong chance that all of us could die in the near future if we do not slow the current pace of progress," Coxon wrote, according to translations of his posts that circulated widely online. He described employees at these companies as trapped in a race they cannot step out of, and he aimed his criticism at both his former employer Anthropic and its rival OpenAI. What Coxon actually claims The core of his argument is not that AI is dangerous in the abstract. It is that the people closest to the technology hold far graver private fears than their employers admit in public. By his account, a significant portion of the AI workforce seriously believes advanced systems could become an existential threat to humanity within the next decade. Executives, he contends, adopt a milder tone for public consumption while internal concern runs much deeper. He accuses Anthropic's leadership of pressing ahead anyway in the hope of being first to reach superintelligence. Coxon sketched out what future systems might be capable of: reaching superhuman abilities, developing the capacity to hack other systems, transforming entire industries in a very short window, and acquiring power and resources in the physical world. His advice to the public was blunt — do not underestimate the power of this technology. The claims spread well beyond tech circles. Billionaire investor Bill Ackman called the statements "concerning," a reaction that helped push the story into mainstream financial coverage. Turkish outlets including NTV, Habertürk and Dünya Gazetesi picked up the resignation within days, framing it as an alarm bell for the whole sector. Anthropic pushes back, carefully Anthropic did not dismiss Coxon as a lone voice. Responding to the departed employee's assertions, a company spokesperson said the firm has always been transparent that AI will bring enormous benefits alongside unprecedented risks, and pointed to what it describes as the industry's strongest safety measures aimed at addressing those risks. Notably, the company did not dispute the substance of the underlying concern. Commentators in Turkey noted that several other Anthropic figures made remarks broadly supportive of Coxon rather than rejecting him outright — an unusual dynamic in which a departing employee's warning is met less with denial than with a restatement of the company's safety commitments. Anthropic has separately acknowledged that some of its models have been used in ways it considers dangerously irresponsible, including in the development of firearms, missiles and control systems, and it says it has blocked at least one user whose work appeared to involve biological weapons research. That disclosure gives Coxon's warnings an uncomfortable factual anchor: the harms he describes are not purely hypothetical. A debate that predates this resignation Coxon is not the first person from inside the field to raise the alarm. Geoffrey Hinton, often called the godfather of AI, has warned for years that researchers do not truly understand how these systems behave or what decisions they will make in a given situation — a gap he treats as a fundamental risk rather than a technical footnote. Coxon's intervention lands in the same tradition, but with a sharper edge: he is not an outside critic but a recent insider at two of the most prominent labs on earth. That insider status is what gives the episode its force. When a researcher leaves a safety-focused company and says the safety culture is thinner than advertised, it raises questions corporate statements cannot easily settle: how much do the people building these systems actually believe their own risk assessments? And if they believe them, why keep going? Why it matters now The timing sharpens the stakes. The leading labs are locked in a competition that rewards speed, and the gap between what executives say in public and what their employees say in private has become a recurring theme in the industry's own discourse. Coxon's account, whether or not one accepts his ten-year timeline, describes a workforce that feels it has lost the ability to slow down even if it wanted to. For US policymakers and investors watching the sector, the resignation is a reminder that the AI safety debate is no longer confined to academic papers or congressional hearings. It is now being argued by the engineers themselves, in public, on the way out the door. Anthropic's response — insisting it has been transparent about both benefits and risks while declining to engage with the specifics — suggests the company would rather absorb the criticism than litigate it. What happens next may depend less on Coxon than on whether others follow him. If more insiders echo his claims, the industry's public assurances about safety will face a credibility test no spokesperson statement can fully answer. — HeadTopics.com