Thalidomide killed thousands, Fukushima zero: where rogue AI harm likely lands

binarybits · x · 2026-09-06

Timothy Lee refines his rogue AI risk analogy: Thalidomide killed thousands while Fukushima killed zero. He expects AI hacking to land somewhere in that rather wide range — from zero-fatality infrastructure incidents to slow-rolling catastrophes noticed and root-caused only after the fact.

The addendum extends his earlier point: rogue AI risk needn't be measured against extinction; mid-scale disasters that still trigger regulatory change are the more useful reference.

Related event: Debating rogue AI risk: Fukushima versus Thalidomide(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →