Anthropic researchers quit with warnings that AI labs are gambling with our lives

Three Anthropic researchers went public with concerns about out-of-control AI, warning that it could destroy humans this decade¹. Researcher Jacob Coxon posted on X after resigning on Tuesday¹. A departing researcher said frontier AI labs are ‘gambling with our lives’^6,7.

The BBC reported a researcher who believes there is more than a 10% chance AI could kill all humans, the latest in a series of increasing warnings about the safety threat posed by the technology². Semafor reported that the researcher who quit warned the industry has not internalised the ‘civilizational stakes’ of rushing towards superintelligence³. Gizmodo carried the claim that the people building AI earnestly believe it could kill us all⁸.

Why it matters to investors

The timing is corporate as well as technical. Daily Sabah reported that the resignation comes as Anthropic prepares for its market debut⁵.

The researcher who quit had previously worked at OpenAI⁶, and Politico reported that the departure came with a dire warning about safety⁴. The Hindu reported the same charge levelled at frontier labs as a group rather than at one company⁷.

The public nature of the break is the change. The researchers went to X and to reporters rather than through internal channels¹, and the BBC treated the episode as the latest step in an escalating series of warnings rather than as an isolated resignation².

The dilemma is industry-wide. Axios reported AI leaders sounding the alarm as their own breakthroughs sharpen a choice between slowing down and risking falling behind, or pressing ahead and risking loss of control⁹.

What to watch

Whether the warnings change disclosure is the first question^1,2. They are being made publicly, by named insiders, rather than raised internally^6,7.

The listing is the second⁵. Safety warnings from named former staff are now part of the public record around that debut^3,4, and the same argument is being made about competitors as well as about one lab^8,9.

The third is whether anything slows. The researchers describe an industry that has not internalised the stakes of the race it is running³, and the leaders of that industry describe a choice they cannot resolve on their own⁹. Nothing in the reporting points to a coordinated pause, only to individuals leaving and saying so publicly^1,6,7.