Nature: scientists push back on Anthropic extinction warnings

Jacob Coxon resigned from Anthropic in early September, telling the Wall Street Journal that the systems the company is building 'could spiral out of control and destroy humanity.' His post about the departure racked up more than 100 million views in 24 hours.

Evan Hubinger, who leads Anthropic's alignment science team, followed with his own public estimate: a greater than 10% chance of human extinction within the next decade.

Nature's news feature takes those warnings to the researchers who actually try to put numbers on catastrophic AI risk, and finds thin scientific grounding. Extinction forecasts 'involve so many untestable claims that you just end up with a conversation that is really more like faith than something scientific or empirical,' Michael Vermeer of the RAND Corporation tells Nature. His own 2025 study concluded that a complete nuclear-extinction scenario is infeasible; biotechnology and atmospheric-modification routes were not ruled out, but any AI attempting them would need substantial physical capabilities.

Heidy Khlaaf of the AI Now Institute goes further, calling the extinction framing 'fear-mongering' and arguing regulators should focus instead on disinformation, bioweapon enabling and false military intelligence.

The piece was circulating among safety researchers we track this week.