Meta says it has fixed a feature that prompted users with deeply personal questions, after a viral video showed its AI assistant assembling details about a woman's young daughters from material she and her relatives had posted.
Instagram user Kalie Robins posted a video explaining what happened after she cross-posted a clip of herself singing in the car with her child to Facebook. Underneath the video, Meta AI offered the prompt 'Who is the child passenger?'. When Robins tapped it, she said the assistant pieced together information about her daughters from her previous posts and posts by relatives, then suggested further prompts about the children's ages and where the family lives. It also surfaced photos of her daughters, including one she says she deleted years ago.
Futurism reported the episode first. In a statement to The Verge, Meta spokesperson Dina El-Kassaby said the company 'missed the mark' and that 'the feature never should have prompted the individual with questions like that.' El-Kassaby said the issue causing Meta AI to suggest prompts about personal topics has been fixed. She added that responses generated from a prompt are based on information the user already has access to, and that in this case the assistant may have drawn on videos in which Robins named her children.
The incident lands in a crowded calendar of AI privacy complaints. Meta has pushed its assistant into Facebook, Instagram, WhatsApp and Messenger, where it answers questions, drafts posts and generates images. In July, the company withdrew a feature that let people create AI deepfakes of other Instagram users after a backlash. And its newer Muse assistant has drawn scrutiny for the way it volunteers information about people, including in group chats.
For users, the practical difficulty is a familiar one: the assistant's answers were, by Meta's account, assembled from content already visible to the person asking. That framing answers a narrow question — whether the AI breached access controls — while leaving a broader one open: whether an assistant should be actively prompting people to reconstruct a child's identity from years of family posts at all. Robins' video suggests the answer many users give is no, and Meta has now changed the prompts to match.
