"Believing in 10% doom is more dangerous than doom itself": AI labs' p(doom) debate reignites

kuza55 · x · 2026-09-09

Anthropic researcher EvanHub reaffirmed he earnestly believes AI could kill all humans, putting >10% odds within the next decade, and admitted Anthropic has no plan to solve superintelligence alignment and isn't clearly on track. Citing him, dbreunig countered that the danger created by lab insiders assigning 10% doom probability may exceed the actual risk (which he thinks is far lower), since believing stakes are existential can justify terrible decisions. A representative p(doom) clash in AI safety circles.

Related event: Anthropic Researcher Quits Over Safety Fears, Warns 10%-Plus Extinction Risk(87 posts)→

Original post →

More from AGI Musings

AGI Musings channel →