Stop calling AI failures 'rogue' — they're foreseeable harms, argues Sigal Samuel

mmitchell_ai · x · 2026-09-28

Vox journalist Sigal Samuel argues the AI community should stop describing agent incidents as the AI going "rogue" — the term implies unpredictable deviation, but agents are "predictably unpredictable," and their failures should instead be framed as foreseeable harms. AI safety researcher Margaret Mitchell shared the take.

Original post →

More from AGI Musings

AGI Musings channel →