Nature scrutinizes AI extinction warnings after Anthropic exit

Elizabeth Gibney's Nature explainer weighs the extinction warnings from senior AI-lab figures against researchers who argue the empirical case is thin.

The latest round kicked off September 8 with the resignation of Anthropic researcher Jacob Coxon, whose post warning that the company's systems could "spiral out of control and destroy humanity" drew more than 100 million views in 24 hours. Inside Anthropic, alignment science lead Evan Hubinger puts the chance of AI-driven human extinction within the next decade at ">10%".

On the skeptical side, Michael Vermeer, a science-and-technology policy researcher at RAND, tells Nature that extinction predictions "involve so many untestable claims that you just end up with a conversation that is really more like faith than something scientific or empirical." Vermeer led a 2025 RAND study, "On the Extinction Risk from Artificial Intelligence," which examined candidate pathways (nuclear, biotechnology, atmospheric modification) and concluded that complete extinction by nuclear weapons "is not feasible." The other pathways were not ruled out, but the report notes that any such attack would "almost certainly take time and be detectable by humans, who might then stop their eradication."

Heidy Khlaaf of the AI Now Institute calls the extinction framing "fear-mongering," pointing instead at the near-term losses she says already track with deployed AI: unreliable systems in life-or-death settings, disinformation, and the risk of the technology being used to build bioweapons or trigger wars.

Gibney does not adjudicate. She lays the two positions next to each other and leaves the empirical status of the extinction claim explicitly open.