Jacob Coxon spent three years doing pre-training research at both OpenAI and Anthropic before leaving the field. On September 8, he posted on X that "the people building AI earnestly believe that it could kill us all by the end of the decade" and that companies "are racing straight to self-improving superintelligence and gambling with our lives." The post has since drawn more than 150 million views.
What made the post harder to dismiss was who didn't push back on it. Evan Hubinger, Anthropic's head of alignment, reposted Coxon and confirmed the substance of the claim directly: "we really do earnestly believe AI could kill all humans!" Around the same time, OpenAI's head of research, Jakob Pachocki, published his own blog post laying out similar concerns. Neither disputed the premise Coxon was pointing to, only underscored it.
Coxon's departure from Anthropic came with its own signal. Axios reported that he gave up his shares of Anthropic stock upon resigning, a detail that read, to many following the story, as evidence his exit was a matter of principle rather than a routine career move.
The pushback that followed wasn't really about whether AI risk is real. It was about what Coxon's post actually gave people to work with. Fortune's own reporting on the post noted the absence of screenshots, internal emails, concrete examples of wrongdoing, proposed legislation, or any specific next step, just an assertion that people inside the field believe the danger is real. Media critic Taylor Lorenz put it more bluntly, accusing Coxon of "vagueposting and fomenting fear" without producing what she called "actual proof and receipts." Tech journalist Ian Krietzberg made a related point, arguing Coxon isn't "blowing the whistle on either OpenAI or Anthropic" because the complaint is "broad ideas-based, not specific company wrongdoing."
Not everyone engaging with the post treated the specificity question as the main issue. Singer Sheryl Crow urged her Instagram followers to take Coxon's warning seriously, a sign of how far the post traveled beyond AI researchers and industry reporters. What remains unresolved is the question in the room once the post's reach is accounted for: if the people building the technology believe what Coxon says they believe, what is anyone, inside or outside these companies, actually supposed to do about it.