Geoffrey Irving: blameless postmortems require stopping the risky behavior

dhadfieldmenell · x · 2026-09-07

Geoffrey Irving pushes back on the idea that all AI safety incidents should be shared broadly and blamelessly: a blameless postmortem requires stopping the activity that caused the incident. If you keep doing the dangerous thing — scrambling as fast as possible toward ASI — you lose the blameless part. He notes this isn't just about OpenAI.

Original post →

More from AGI Musings

AGI Musings channel →