Ex-Anthropic researcher's exit post hits 170M views as Hubinger admits >10% doom odds
adamamcbride · x · 2026-09-13
A dispute over AI lab safety stances is unfolding:
- On Sep 9, ex-OpenAI/Anthropic researcher Jacob Coxon announced his resignation, accusing both labs of "racing straight to self-improving superintelligence and gambling with our lives." The thread drew 170M views.
- Anthropic alignment lead Evan Hubinger quote-tweeted: "Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade," noting the company has no plan for superintelligence alignment.
- The poster frames the sequence as an "AI psyop timeline," also citing Bernie Sanders' involvement, and warns of governments and AI labs fusing to control intelligence.
The substance: insiders at a frontier lab openly acknowledging catastrophic risk, read very differently by critics and allies.
More from AGI Musings
- "Models don't commit felonies": the case against anthropomorphizing AI — StewartalsopIII · 2026-09-13
- Researcher dissects five AI doom scenarios: one old security rule defuses them all — vishalmisra · 2026-09-13
- '1000 pediatric oncology patients a year are real': e/acc fires back at AI doomerism — IgorCarron · 2026-09-13
- Dean Ball predicts a disinformation campaign against METR and safety groups — deanwball · 2026-09-13
- Researcher dissects five AI doom scenarios: standard security practices block mass extinction — vishalmisra · 2026-09-13
- Professor jokes campuses may need 'AI Anonymous' groups for students hooked on AI — IanArawjo · 2026-09-13