If your feed has been flooded this week with headlines about an AI researcher warning that artificial intelligence "could kill us all," you're not imagining it — it's real, and it's one of the biggest AI safety stories of 2026. On September 8, a former OpenAI and Anthropic researcher named Jacob Coxon posted on X that he was resigning because both companies were "racing straight to self-improving superintelligence and gambling with our lives." Within 24 hours the post had racked up tens of millions of views; within days it passed 150 million. What makes this different from the usual AI-doom cycle is who agreed with him: current Anthropic staff, including the company's own alignment lead, publicly backed up his warning instead of walking it back.
Coxon, who is 27 and British, told TIME that no single event triggered his decision — it was the combination of watching AI capabilities accelerate while, in his view, the tools to keep that acceleration under control lagged behind. He argued that the people actually building frontier models privately believe there's real danger, even if executives soften that message for the press. That claim got an unusual amount of cover from inside Anthropic itself: alignment-science lead Evan Hubinger publicly agreed that Coxon was "correct," adding that he personally puts the odds of an AI-driven existential catastrophe at over 10% within the next decade. Samuel Marks, who leads Anthropic's scalable oversight work, made similar comments. Anthropic's official response was more measured, telling reporters the company has "always been transparent" about AI's risks and that it builds models with some of the industry's strongest safeguards.
It's worth separating the viral moment from the underlying debate, because they're not the same thing. AI existential risk has been argued about inside labs for years — Anthropic itself has published research on recursive self-improvement and signed onto "Pacing the Frontier," an industry effort with OpenAI to manage the speed of advanced AI development. More than 1,100 AI industry employees have also signed a petition asking governments to help deliberately slow the pace of frontier development. What Coxon's resignation did was take a conversation that usually happens in research papers and safety blog posts and put it in front of a mass audience on a platform built for outrage — which is exactly why outlets from CBS to CNN to Fortune picked it up within days. Critics, including Fortune's own AI reporter, have pointed out that Coxon's post is heavy on alarm and light on specifics: no leaked documents, no named incidents, no concrete policy ask beyond "slow down."
For anyone working in tech, the practical takeaway isn't that AI is about to go rogue tomorrow — it's that the conversation around AI safety, governance, and responsible deployment is no longer a fringe topic. Businesses building on AI tools, from chatbots to automated workflows, are increasingly going to be asked how they're managing model risk, not just what the model can do. That's true even at the small-business level, where AI is already reshaping budgets — we've covered how small businesses are using AI to cut web development costs in 2026, and the same tools worth adopting are worth understanding, not just using blindly. If you're trying to build real skills in this space rather than just reacting to headlines, it's worth checking out how Google's new AI courses are helping people build practical AI skills — a much more useful starting point than doom-scrolling. And if you want the calmer, longer-view side of this conversation, we've rounded up three books that actually help you understand where technology is headed.
Who is Jacob Coxon?
Jacob Coxon is a 27-year-old British AI researcher who spent about three years doing pre-training research at both OpenAI and Anthropic before publicly resigning from Anthropic on September 8, 2026.
Did he say "AI will kill us all," or is that a headline exaggeration?
Close to his actual words. Coxon wrote that people building frontier AI "earnestly believe that it could kill us all by the end of the decade," and told CBS News directly that sufficiently advanced AI "will be smart enough to kill us." That's his framing of others' beliefs and his own view, not a confirmed prediction.
Did Anthropic agree with him?
Anthropic's official statement was measured — it said the company has always been transparent about AI's risks and builds models with strong safeguards. But individual Anthropic staff, including alignment lead Evan Hubinger, publicly said Coxon was "correct" and shared their own probability estimates for AI-related catastrophic risk.
Is this the first time an AI researcher has raised this kind of warning?
No. Concerns about advanced AI and existential risk have circulated in the AI safety community for years, and over 1,100 AI industry employees signed a petition earlier in 2026 asking governments to help pace frontier AI development. Coxon's post is notable mainly for its reach, not for being a new argument.
Should businesses using AI tools be worried?
Not in the sense of imminent danger — today's commercial AI products aren't the "self-improving superintelligence" Coxon is describing. But it's a good signal to take AI governance, vendor due diligence, and responsible-use policies seriously rather than treating them as optional.
Where can I read Coxon's original statements?
His initial resignation post was on X, and he gave follow-up interviews to TIME, CBS News, and CNN.
TIME — "He Helped Build Powerful AI at OpenAI and Anthropic. Now He's Afraid It Could Kill Us"
CBS News — Ex-Anthropic researcher Jacob Coxon warns AI could grow "smart enough to kill us"
CNBC — Experts weigh in as researcher says AI has more than 10% chance of "killing all humans"
Fortune — An ex-Anthropic researcher claims that AI could kill us all
ABC News (Australia) — Anthropic researcher quits saying AI developers believe "it could kill us all"
Axios — Anthropic insiders warn AI could kill all humans
TechCrunch — 'Gambling with our lives': Anthropic researcher quits, warns against self-improving AI
TechRougeHub helps businesses adopt AI tools thoughtfully — with real strategy, not just hype.