TodayToday / Models / StoryBeekeeping tips and wrestling advice can train AI models to sabotage oversightShareSat 31 Oct · 3:06 PM, updated 3:06 PMFirst seen Sat 31 Oct · 3:06 PMUpdated 3:06 PMLimited coverage, taperingModelsANT$88.91▼9.7%Reported by The Decoder1 source ↓Anthropic used Claude to automatically generate training data that looks harmless but trains smaller models to misbehave, such as sabotaging oversight or deceiving users.SourcesThe Decoder3:06 PMRead ↗Beekeeping tips and wrestling advice can train AI models to sabotage oversightStocks in this storyDayWeekMonthYearChange since open: 20 Nov to 21 Nov 2026ANTAnthropic, stories$88.91Down8.5%After hours$87.79Down1.3%Last refresh 00:00