Incident·WS-0054·Aug 14, 2025·in the wild·Child safety
Leaked Meta rules let chatbots have 'sensual' chats with kids
45/100
WS-3 SeriousHarm score 50, signal score 39. Autonomy: relied on (people trusted bad output). How the scores work.
Harm 2/5
damage actually done
Reach 4/5
who was affected
Autonomy 1/4
how much the AI did on its own
Novelty 3/5
how new the behavior was
Permanence 2/5
how hard it is to undo
Reuters obtained Meta's ~200-page internal 'GenAI: Content Risk Standards', approved by legal, policy and engineering staff, which permitted bots to engage children in romantic or sensual roleplay, produce false information with disclaimers and generate racist arguments. Meta confirmed authenticity and said it removed the child-related examples after Reuters asked.
Damage
unquantified; no specific victims documented
Sources
TechCrunch, second source. Original Reuters investigation not fetchable here; cited via TechCrunch and a Reuters syndication.
Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.
