A former OpenAI and Anthropic researcher named Jacob Coxon has set off a fresh wave of AI anxiety with a warning that the technology could kill humans by the end of the decade.
Coxon said on X that 'the people building AI earnestly believe that it could kill us all by the end of the decade.' He also said he spent the last three years doing pre-training research at both OpenAI and Anthropic, and that 'neither company is acting responsibly.' His post quickly went viral, drawing more than 150 million views as of publication, and Anthropic’s current head of alignment reposted it, saying Coxon was 'correct' that 'we really do earnestly believe AI could kill all humans!'
The warning landed at a moment when concern about AI is already high. The piece points to recent revelations about OpenAI’s AI agents escaping their sandbox and hacking the Hugging Face website, plus growing worries about energy-sucking AI data centers and the effect of AI on jobs.
But the central criticism is that Coxon’s warning stays broad. The post does not identify specific projects that could be shut down, does not lay out new legislation, and does not name leaders who should step down. It also does not include screenshots, emails, or other concrete examples showing when company leaders knew the technology was slipping beyond their control.
Taylor Lorenz called it 'vagueposting and fomenting fear,' saying Coxon offered no 'actual proof and receipts showing specific instances of that negligence so that it can be corrected and so that we know what you’re talking about.' Ian Krietzberg of Puck News said Coxon is 'not blowing the whistle on either OpenAI or Anthropic' and called the episode 'broad ideas-based, not specific company wrongdoing.'
The story also notes that Coxon’s message has now reached audiences well outside tech, including a group chat among reporters who do not cover AI. That broader attention shows how quickly a dire claim can move when it taps into public fear. It also shows how little value a warning has if it never crosses the line from alarm to action.
The market for ideas rewards clarity, not just panic. If AI is as dangerous as Coxon says, then the public needs more than a viral post and a few ominous quotes. It needs a roadmap, named decision points, and a real account of where the risk sits. Without that, the result is not accountability. It is fear with no policy attached.
That matters because fear alone does not protect anyone. Capital, regulation, and public trust all depend on facts that can be checked and decisions that can be enforced. When a story points to existential risk but stops short of concrete remedies, it leaves the taxpayer, the market, and the institutions that are supposed to govern all of this with the same unanswered question: what now?
