Landmark AI researcher survey: most experts see 10% chance AI destroys humanity
DavidSKrueger · x · 2026-09-15
KatjaGrace's team released what's billed as the biggest and best survey of AI researchers to date.
Key finding: most surveyed AI researchers assign roughly a 10% probability to AI destroying or disempowering humanity. The worry isn't limited to a few insiders at AI companies — it's a mainstream view among researchers.
The thread also cites Jacob Coxon's departure warning from Anthropic and colleague Evan's ">10% chance" estimate that AI would "kill all humans," noting such views are not unusual in the field.
More from AGI Musings
- AI extinction risk estimate rises to 10% in expert survey; debate over probability framing — NathanpmYoung · 2026-09-15
- Ryan Lowe: AGI needs coordination infrastructure — which is just institutions by another name — ryan_t_lowe · 2026-09-15
- Microsoft's blunt 'It is not conscious' won't settle the AI consciousness debate — rgblong · 2026-09-15
- AI masters the 'last mile' of idea polishing — and will keep going until humans take the backseat — andrew_n_carr · 2026-09-15
- Dario: Anthropic controlling AGI would be like Hitler getting the atomic bomb first — teortaxesTex · 2026-09-15
- Anthropic's AI economics report criticized for ignoring economic feedback loops — ziv_ravid · 2026-09-15