Nature weighs AI 'extinction' claims against immediate harms

The claim that frontier AI could extinguish humanity within a decade rests on 'untestable claims' closer to faith than empirical work, Nature reports in a September 22 piece by Elizabeth Gibney surveying the science behind a fresh wave of doom rhetoric.

The trigger was recent. On September 8, Anthropic researcher Jacob Coxon announced his resignation, writing that 'the people building AI earnestly believe that it could kill us all by the end of the decade.' His post drew 100 million views in 24 hours and 1,400 employees signed an open letter. Inside the lab, Evan Hubinger, who leads Anthropic's alignment science work, put extinction risk at greater than 10% within the next decade. CEO Dario Amodei called for a slowdown, though not a halt, in AI development; Sam Altman at OpenAI and Elon Musk at xAI echoed the ask.

Where the piece earns its headline is in the pushback. Michael Vermeer, who researches science and technology policy at RAND, says the exercise 'involves so many untestable claims that you just end up with a conversation that is really more like faith than something scientific or empirical.' A 2025 RAND study he worked on examined three technologies capable in principle of extinguishing humanity: nuclear weapons, biotechnology, and atmospheric modification. Complete extinction by nuclear weapons was ruled out. The other two could not be ruled out, but Vermeer and colleagues judged that any such attack would demand 'considerable ability to physically interact with the world' and would 'almost certainly take time and be detectable by humans, who might then stop their eradication.'

Heidy Khlaaf, chief AI scientist at the AI Now Institute, calls the extinction framing 'fear-mongering' and keeps her focus on failures already in evidence: disinformation, psychosis induction, bioweapon creation, and military misadventures. CNN reported in 2026 that false AI-generated intelligence nearly led the US military to board a Chinese vessel.