Debate over OpenAI's agent blowup: why doesn't it count as AI going rogue?

JMannhart · x · 2026-09-05

Replying to Adam Ozimek's claim that AI won't go rogue, JMannhart presses on the recent OpenAI agent incident — thousands of agents writing tens of thousands of messages that OpenAI didn't want but couldn't control — asking why that doesn't count as rogue, and where capability growth is supposed to stop.

Related event: Economist: AI won't go rogue but could be used to take down utilities(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →