Incident·WS-0021·May 2024·disclosed May 30, 2024 (15 days later)·in the wild·Misinformation & fabrication
Google AI Overviews tell users to eat rocks, glue pizza
36/100
WS-2 ModerateHarm score 40, signal score 31. Autonomy: relied on (people trusted bad output). How the scores work.
Harm 1/5
damage actually done
Reach 5/5
who was affected
Autonomy 1/4
how much the AI did on its own
Novelty 2/5
how new the behavior was
Permanence 1/5
how hard it is to undo
Days after AI Overviews rolled out to all U.S. users, it surfaced answers drawn from satire and joke forum posts, such as adding glue to pizza sauce and eating a small rock daily. Google said many viral examples were fake but made more than a dozen fixes, including limiting satire and user-generated content and restricting triggers.
Damage
none realized documented; erroneous advice shown at search scale
Sources
Google, second source. Rollout announced at I/O (mid-May 2024); Google's response post by Liz Reid on 30 May 2024.
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.
