An Anthropic researcher resigns with a public warning about AI risk

Jacob Coxon, a researcher at Anthropic who had previously worked at OpenAI, resigned and made his departure public, telling colleagues in a company Slack message — later posted to X, where it drew tens of millions of views — that the leading AI labs were “racing straight to self-improving superintelligence and gambling with our lives.” He put the probability of human extinction from AI at more than 10% within the decade and said he could no longer reconcile that belief with continuing to work on frontier capabilities.

According to NBC News, which reported the Slack message and a same-day interview, Coxon argued that the industry’s competitive dynamics made it impossible for any single lab to slow down unilaterally, and that the pace of progress had outrun the field’s ability to ensure models remained under human control. He framed his resignation less as a criticism of Anthropic specifically than as a protest against the trajectory of the industry as a whole.

The episode was notable because a senior colleague corroborated rather than contested him. Evan Hubinger, who leads alignment work at Anthropic, responded publicly that he shared Coxon’s rough estimate of more than 10% extinction risk within a decade and acknowledged the field had no worked-out plan for aligning superintelligent systems. The exchange fed a wider run of public risk statements from lab insiders that fortnight, including OpenAI chief scientist Jakub Pachocki’s “An Alien Mind” essay and Dario Amodei’s call to “pace the frontier” — a cluster that observers read as senior researchers becoming more willing to voice concerns publicly that had previously stayed internal.