Anthropic's Evan Hubinger: >10% chance AI kills all humans within a decade
CtrlAltDwayne · x · 2026-09-10
Responding to criticism, Anthropic's Evan Hubinger said he earnestly believes AI could kill all humans, personally estimating >10% probability within the next decade. He admitted Anthropic is trying its best but has no plan to solve alignment for superintelligence and isn't clearly on track. Commenters mocked the response's AI-writing tells like em-dashes.
Related event: Anthropic Researcher Resigns Warning of AI Risks(205 posts)→
More from AGI Musings
- Anthropic models AI's impact on jobs through 2030 using nursing as a case study — HealthcareAIGuy · 2026-09-10
- Prediction: AI disclosure statements in papers will be quaint by 2028 — Aaroth · 2026-09-10
- A Redditor drafts six future-proof AI constitution principles, crowdsourcing critique — Alex__007 · 2026-09-10
- Slow-takeoff champion Paul Christiano now says fast takeoff may be months to years away — DavidSKrueger · 2026-09-10
- Self-described AliExpress employee resigns, warns of race to self-improving supply chains — haydendevs · 2026-09-10
- AI safety debate continues: from mirror life to cognitive decline, the full risk spectrum — BlackHC · 2026-09-10