AI Researchers Back Extinction Warning as Pentagon Rejects It
AI researchers back an extinction warning as the Pentagon rejects it
An AI extinction warning from a departing Anthropic (get the details) researcher drew a wave of responses this week. Scientists at Anthropic and OpenAI backed the claim, while a top Pentagon official dismissed it. The row reached Washington and Anthropic’s planned IPO.
September 10, 2026 – 9:24 pm
Credit: jackpress / Shutterstock
A warning from Jacob Coxon, a former Anthropic researcher, has sparked debate. In a post on X, he claimed AI firms were “racing straight to self-improving superintelligence and gambling with our lives”. The post went viral, reaching over 150 million views according to Bloomberg.
Coxon, who left Anthropic before his equity vested, said he had seen enough during his three years in AI research at both OpenAI and Anthropic to believe that AI could pose an existential threat to humanity by the end of the decade. His concerns were echoed by several colleagues and rivals, including Evan Hubinger, Samuel Marks, and Paul Christiano.
Christiano, formerly head of safety at the US Commerce Department, expressed a similar view, stating that recent progress in AI had led him to see a meaningful risk. Geoffrey Hinton, often referred to as a godfather of AI, agreed, estimating a 10% chance of human extinction as "not unreasonable". OpenAI itself announced Christiano’s appointment to its board.
However, not all were convinced. Emil Michael, the US Defense Department’s chief technology officer, dismissed the warning at a defense industry conference in Washington, characterizing it as a "doom loop" and "all the fears that had coalesced in one well-written tweet". He expressed confidence in the free market, industry collaboration, and government engagement to mitigate potential harms.
This debate comes after an open letter signed by 1,400 AI employees urging the US government to pace automated AI development.