Meta Fixes AI Prompt That Surfaced Child Data From 300,000-Like Instagram Video
Updated
Updated · Yahoo · Sep 9
Meta Fixes AI Prompt That Surfaced Child Data From 300,000-Like Instagram Video
1 articles · Updated · Yahoo · Sep 9
Summary
Meta said it fixed an Instagram AI prompt after Utah mother Kalie Robins found it asking “Who’s the child passenger?” on her video and then surfacing personal details about her children and past addresses.
Those prompts appeared to pull from posts across social media, including relatives’ Facebook photos, deleted family images and clues that Robins said could point to her children’s school.
Meta said the feature was meant to help users learn more about posts they already could access, but acknowledged Robins should never have received questions like that about a minor.
Robins’ follow-up warning about the incident drew more than 300,000 likes and prompted her to remove identifiable images of her children and ask friends and family to do the same.
The episode adds to scrutiny of Meta’s AI tools after recent complaints about inconsistent AI labeling, a rolled-back image feature and a newly announced $17 billion settlement over teen app use.