A researcher’s resignation puts AI extinction risk in front of Congress

An Anthropic researcher said he is resigning from the company over concerns that the artificial intelligence firm and its competitors are not acting responsibly². Fortune reported the departure under his description of the industry as gambling with our lives¹. The researcher believes there is a more than 10% chance that AI could kill all humans, the BBC reported, calling it the latest in a series of increasing warnings about the safety threat posed by artificial intelligence³.

The Pentagon rejected the warning. Emil Michael, the Defense Department’s chief technology officer and its top research and engineering official, dismissed the former Anthropic engineer’s argument⁴. Michael also said Anthropic technology has been removed from 90% of classified military AI work⁵.

David Sacks called for Anthropic’s initial public offering to be paused until whistleblower claims are investigated, putting the company’s listing in question⁷.

Why it matters to investors

The dispute has moved from laboratories into procurement and capital markets in a single week. Removal from the overwhelming majority of classified military AI work is a revenue and positioning event, not only a reputational one⁵, and a call to pause a listing bears directly on the exit path for private AI valuations⁷.

Legislative risk is the second channel. Warnings from company insiders that AI could end humanity in the next decade have members of Congress frantically seeking answers and trying to assemble a response⁶. Lawmakers are at the concepts of a plan stage, with congressional leaders showing little apparent urgency to mobilise⁶.

The pressure is bipartisan, which matters for the odds of something passing. We should absolutely act, and act with urgency, Rep. Johnny Olszewski, a Maryland Democrat, told Axios⁶. Rep. Anna Paulina Luna, a Florida Republican, wrote that Congress needs to convene a special session on AI and what the future of the country looks like⁶.

What to watch

The first test is whether the congressional reaction converts into a bill or stops at hearings, given the gap between rank-and-file urgency and leadership inertia described by Axios⁶.

The second is the defence relationship. Whether other suppliers take the classified work vacated by Anthropic⁵, and whether Michael’s public dismissal of the extinction argument sets the department’s procurement posture⁴, will shape which vendors hold federal AI contracts.

The third is the listing. A pause called on the basis of whistleblower allegations⁷ would be the first instance of safety claims directly delaying an AI public offering, and the precedent would apply well beyond one company. The warnings themselves are not slowing: the resignation is described as the latest in a lengthening series^2,3.