# 100 Million Views: Anthropic Researcher Quits With Stark Warning That AI Could Kill Humanity > Source: > Published: 2026-09-10 13:20:41+00:00 # 100 Million Views: Anthropic Researcher Quits With Stark Warning That AI Could Kill Humanity ## Anthropic researchers warn that superintelligent AI could put humanity's survival at risk An Anthropic AI researcher has resigned with a warning that has sent shock waves through the artificial intelligence industry: people building advanced AI genuinely believe it could one day kill humanity. [Jacob Coxon announced his resignation](https://www.ibtimes.co.uk/ai-researcher-jacob-coxon-quits-industry-warns-superintelligence-risks-1818907) on X on Tuesday, saying he had spent the previous three years doing pretraining research at both OpenAI and Anthropic. His warning quickly went viral, reaching more than 110 million views and triggering an extraordinary response from researchers still inside Anthropic. But the most significant response came from one of his Anthropic colleagues. ## Anthropic Insider Makes Chilling Admission Evan Hubinger, Anthropic's Alignment Science Lead, responded directly to Coxon with a blunt assessment of the danger. 'We really do earnestly believe AI could kill all humans!' Hubinger wrote, adding that he personally believed the chance was greater than 10 per cent within the next decade. That figure is Hubinger's personal assessment, not a measured scientific probability or an established consensus on the likelihood of human extinction. His next admission was arguably even more significant. Hubinger said Anthropic was trying its best, but did not yet have a plan to solve alignment for superintelligence and was not clearly on track to do so. Alignment refers to the challenge of ensuring increasingly powerful AI systems continue to behave in ways compatible with human goals and safety. ## Why Jacob Coxon Walked Away Coxon accused Anthropic and OpenAI of racing towards self-improving superintelligence without adequately addressing the dangers. 'They are racing straight to self-improving superintelligence and gambling with our lives,' he wrote. His concern is not that today's chatbot will suddenly become a science-fiction villain. The focus is on future systems that could contribute to building more capable AI, potentially creating a cycle of rapid improvement that becomes increasingly difficult for humans to understand or control. Coxon also alleged that executives and senior researchers privately express greater fear about AI than they acknowledge publicly. Those comments remain his allegations and have not been independently established. ## Another Anthropic Researcher Raises the Alarm Coxon's warning was not isolated. Samuel Marks, Anthropic's scalable-oversight lead, said in a personal capacity that AI developers believe their technology could cause human extinction or similarly severe outcomes. He added that more senior employees tend to be more concerned. Marks also highlighted the uncomfortable reason the race continues. Developers may fear that slowing down could allow a rival to build powerful systems first and deploy them less safely. That leaves the industry facing a stark dilemma: slow down and risk falling behind, or keep advancing while the technology becomes harder to control. ## The Risk Is About Future AI Hubinger has stressed that the risk from current models is low. His concern is what could happen if superintelligence emerges through recursive self-improvement, a process in which an AI system could help develop increasingly capable successors. That distinction matters. The debate is largely about future frontier systems, not evidence that today's mainstream AI models are capable of wiping out humanity. Anthropic has also maintained that it is pursuing safety measures. In a recent assessment, the company said the risk from current models was low while examining the potential for more concerning misalignment in future, more capable systems. ## Experts Clash Over the AI Apocalypse Not everyone accepts the most dramatic forecasts. Luke Stark, an assistant professor at Western University who studies the history and ethics of computing, including AI systems, described the 10 per cent estimate as 'science fiction'. He argued that attention should also focus on harms already emerging from AI, including data-centre expansion, environmental costs and growing corporate power. David Krueger, an assistant professor at the University of Montreal and a core academic member of Mila, takes a far more serious view. He has argued that AI risks could be more severe than publicly acknowledged and called for an immediate, indefinite international moratorium on frontier AI development. The disagreement leaves a striking contradiction at the heart of the industry. Anthropic is known for its focus on AI safety, yet one of its senior alignment researchers says the company has not solved the problem of controlling future superintelligence and is not clearly on track to do so. Coxon's resignation has therefore become more than one researcher's protest. It has exposed a question that is becoming increasingly difficult for the AI industry to avoid: If the people building the technology believe it could kill all humans, how far should the race be allowed to go before safety catches up? © Copyright IBTimes 2026. All rights reserved.