Anthropic researcher: >10% chance AI kills all humans within a decade, no alignment plan yet
DKokotajlo · x · 2026-09-10
AI-jobs researcher Molly Kinder describes how weeks of alarming signals — the Hugging Face incident, critical METR eval reports, and pleas from inside the labs — shifted her from passively accepting x-risk talk to genuine fear.
The core quote comes from Anthropic's Evan Hubinger: he and colleagues 'earnestly believe AI could kill all humans,' and he personally puts the odds at >10% within the next decade. He says Anthropic is trying its best, but has no plan to solve alignment for superintelligence and is 'not clearly on track.'
The thread captures a visible mood shift among insiders, from abstract concern to personal dread.
More from AGI Musings
- Michael Black on academic CV research's role in the age of powerful large models — CSProfKGD · 2026-09-10
- AI-run interviews reveal a split: childfree cite freedom, would-be parents cite cost — soumitrashukla9 · 2026-09-10
- Cambridge prof David Krueger puts AI catastrophe risk above 50%, says everyone is understating it — KatjaGrace · 2026-09-10
- MIT Schwarzman College pilots program to help faculty teach AI across disciplines — nordicinst · 2026-09-10
- Engineer deploys hundreds of parallel AI agents to work on a type 1 diabetes cure — Scobleizer · 2026-09-10
- Proactive agent Muse remembers daughter's 8th birthday and offers to plan the party — altryne · 2026-09-10