Meta says it has changed its AI assistant after a suggested prompt led the system to assemble personal information about a user’s young daughters. The company acknowledged that the feature “missed the mark” and said it should not have asked questions of that kind.
The incident began when Instagram user Kalie Robins cross-posted a car video to Facebook. A suggestion beneath it asked, “Who is the child passenger?” After she selected the question, Robins says Meta AI combined information from her posts and relatives’ posts, proposed further questions about her children’s ages and home, and displayed photos of the girls—including one she believed had been deleted years earlier.
Meta says it fixed the issue that generated prompts about personal topics. The company also says responses use information the requesting user can already access, which may include material shared elsewhere across its services. That explanation addresses permissions but not the discomfort created when software connects scattered details into a ready-made profile.
Meta AI is integrated into Facebook, Instagram, WhatsApp and Messenger. The episode illustrates a privacy distinction that access controls alone do not resolve: individual facts may each be visible, while automatically collecting and presenting them—especially about children—can reveal far more than any single post.