WorkSeptember 9, 2026via The Decoder
Anthropic scientist puts the odds of AI destroying humanity above ten percent this decade
Why it matters
A senior AI safety researcher at a frontier lab is publicly stating extreme-tail-risk odds, while a colleague departs citing existential risk concerns. This is a people move + safety commentary with policy/societal implications that practitioners and AI-policy watchers will debate.
Key signals
- Evan Hubinger (Anthropic): >10% probability of misaligned superintelligent AI causing human extinction within this decade
- Jacob Coxon: former pretraining researcher at OpenAI and Anthropic, quit citing extinction risk concerns
- Coxon accuses both OpenAI and Anthropic of knowingly risking human extinction
- Published September 9, 2026
- Evan Hubinger (Anthropic) estimates >10% probability of misaligned superintelligent AI causing human extinction within this decade
- Jacob Coxon (former pretraining researcher at OpenAI and Anthropic) has quit, accusing both companies of knowingly risking human extinction
- Published 2026-09-09
The hook
Anthropic researcher Evan Hubinger puts odds of AI-caused human extinction above 10% this decade—and a former pretraining scientist just quit over extinction risk.
Jacob Coxon, a former pretraining researcher at OpenAI and Anthropic, has quit and accuses both companies of knowingly risking human extinction. Anthropic colleague Evan Hubinger puts the odds of a misaligned superintelligent AI wiping out humanity within the next decade at more than ten percent.