Researcher resigns from Anthropic warning of AI extinction risk

reported by 5 outlets· 15 articles · confidence: low · first seen 2026-09-09 09:56 UTC

What this means for you

Nothing to act on today: no product, price, policy or deadline follows from this. The numbers quoted — more than 10% odds, extinction by 2030 — are individual researchers' stated beliefs, not measurements, and the reports give no method behind them. The thing to watch is whether lawmakers' criticism becomes legislation.

An Anthropic researcher has resigned with a public warning that self-improving AI could kill everyone, Ars Technica reports. The BBC puts one researcher's estimate of that risk at more than 10%; the Guardian reports researchers saying human extinction is possible by 2030. Lawmakers then criticised AI companies, and the Guardian reports Elon Musk dismissed the warnings as a "psyop". Further Anthropic staff added warnings, and a former employee told the BBC that AI staff are frightened. The reports differ on whether one researcher or several are involved.

Key facts

  • ·An Anthropic researcher resigned with a warning that self-improving AI could kill everyone, according to Ars Technica. source
  • ·The BBC reports a figure of more than 10% chance that AI could kill all humans, attributed to an Anthropic researcher. source
  • ·The Guardian reports Anthropic researchers saying AI could cause human extinction by 2030. source
  • ·Lawmakers criticised AI companies after the researcher's warning, the Guardian reports. source
  • ·Elon Musk dismissed the warnings as a "psyop", according to the Guardian. source
  • ·A former Anthropic researcher told the BBC that AI staff are frightened for humanity's future, in a piece published 13 September 2026. source

What the sources say

  • Ars Technica AI — Reports the resignation and the departing employee's warning about AI that improves itself.
  • TechCrunch AI — Podcast asking why the warning landed when it did.
  • The Verge AI — Notes that several employees, not one, are raising the concern.
  • BBC Technology — Supplies the quantified risk estimate and attributes it to an Anthropic researcher.
  • The Guardian AI — Sets the warnings alongside other expert and political voices, questioning how much weight to give them.
  • The Guardian AI — Puts a ten-year deadline on the extinction claim.
  • The Guardian AI — Covers the political reaction from legislators to the AI companies.
  • The Guardian AI — Adds more employee warnings and Musk's rejection of them.
  • The Guardian AI (opinion) — Column arguing the claims deserve attention whatever their tone.
  • BBC Technology — Follow-up interview on how staff at these companies feel about the work.

Sources

The original reporting. Follow these — they did the work.

Related stories

← the wire