Meta Removes AI Prompts Asking About Children and Home Locations

Meta says it has fixed Meta AI prompts that asked users about a child in a post and a person’s home location after a parent’s viral video showed how the system could assemble family details from years of online material. The change stops those suggested questions; Meta says answers can still draw on information users already have access to.

Kalie Robins publicized the episode in an Instagram video posted September 2. She said Meta AI suggested asking “Who’s the child passenger?” after she cross-posted a video of herself singing in a car with one daughter to Facebook. After selecting prompts, Robins said, the tool drew together information from earlier posts and other profiles.

Meta acknowledged that it had “missed the mark” by suggesting personal questions about minors and location. It said Meta AI will no longer make inappropriate personal suggestions on users’ posts. The company describes the feature as a way to help people find more information about posts or topics that interest them.

One part of the episode remains disputed. Robins said the tool surfaced a photo of her children that she had deleted years earlier. Meta denied that account, telling CNET the image was deleted shortly before Robins made her video and that a temporary bug briefly kept it visible. Robins told CNET that the broader problem was not one image, but the AI compiling information about her children from multiple sources.

Robins’ video had drawn more than 308,000 likes when CNET published its report, alongside parental concern about sharing children’s images online. In a later post, she welcomed Meta’s acknowledgement but asked what the company had fixed, which safeguard had failed and what protections now apply when Meta AI encounters information involving a child.