Anthropic researcher: >10% chance AI kills all humans within a decade, alignment unsolved
EvanHub · x · 2026-09-09
- Anthropic researcher Evan Hubinger says AI builders earnestly believe AI could kill all humans — it's not marketing. He personally estimates >10% probability within the next decade.
- He admits Anthropic is trying its best, but there is not yet a plan to solve alignment for superintelligence, and it's unclear the current path leads there.
- The quoted post notes executives and senior researchers hedge in public but express the same fear privately; no other human activity poses this level of danger.
More from AGI Musings
- Pedro Domingos: OpenAI and Anthropic are dishonestly redefining AGI — pmddomingos · 2026-09-09
- Grady Booch: technology is materially transforming the form of law firms — Grady_Booch · 2026-09-09
- One-liner alignment question: why would a smarter species ever care about a weaker one? — petergyang · 2026-09-09
- AI shifts the founder's job from translating ideas into deciding what matters — alexmacgregor__ · 2026-09-09
- OpenAI accused of scooping math researchers who used its models for their work — ns123abc · 2026-09-09
- If AGI is defined by outcomes, we may already be there — sebnadeau · 2026-09-09