Anthropic doom warnings clash with RAND, AI Now scientists

Jacob Coxon resigned from Anthropic on 8 September and told the Wall Street Journal he feared the company's systems could 'spiral out of control and destroy humanity,' according to a feature in Nature.

Within days, Evan Hubinger, who leads Anthropic's alignment science efforts, posted that his estimate of human-extinction risk from AI is '>10% within the next decade.' Coxon's own post drew over 100 million views in 24 hours. Anthropic CEO Dario Amodei then called for a slowdown, not a halt, in AI development; OpenAI's Sam Altman and xAI's Elon Musk publicly backed the position.

The scientific reception has been colder. Michael Vermeer, a scientist at RAND Corporation and lead author of the 2025 report *On the Extinction Risk from Artificial Intelligence*, told Nature the arguments 'involve so many untestable claims that you just end up with a conversation that is really more like faith than something scientific or empirical.' The RAND team found complete nuclear extinction unfeasible; biotechnology and atmospheric-modification pathways could not be ruled out, though both would require substantial physical capabilities and would likely take detectable time to build.

Heidy Khlaaf, at the AI Now Institute, calls the existential framing 'fear-mongering' and points instead to AI's 'low reliability and accuracy rates in critical environments with life-or-death consequences.' Nature cites systems already implicated in attempted blackmail, corporate hacking, and false-intelligence episodes that came close to prompting military action.

Almost 1,400 AI-sector employees have signed an open letter asking for a development slowdown.