Anthropic's Evan Hubinger says AI could kill all humans with >10% odds this decade
wfithian · x · 2026-09-10
Anthropic researcher Evan Hubinger publicly stated that he earnestly believes AI could kill all humans, putting his personal odds above 10% within the next decade, and admitted Anthropic has no plan to solve alignment for superintelligence and isn't clearly on track.
Quoting user littIeramblings adds that as of 2022 Evan already assigned 80% to existential risk from AI, and argues the '>10%' lower-bound framing undersells expert alarm—ranges like 10-20% would communicate it better.
Related event: Seven AI Insiders Warn of Extinction Risk Within a Decade(4 posts)→
More from AGI Musings
- Gary Marcus mocks AGI arguments that rely on 'AI will think of things we can't imagine' — GaryMarcus · 2026-09-11
- 30-year technologist slams AI doom activists: record safety is historic — robleclerc · 2026-09-11
- Markets underprice AGI and overprice stagnation, argues Ronald Qiao — abhiadesai · 2026-09-11
- MILA's Irina Rish: AGI is practically here, ahead of my 7-year CERC schedule — irinarish · 2026-09-11
- Proposing 'Swarm Studies': a research roadmap for autonomous agent swarms — soumitrashukla9 · 2026-09-11
- AI risk isn't about intelligence: 'the action space IS the risk' — josh_wills · 2026-09-11