Meta changes chatbot suggestions after viral video exposes invasive questions about children

Meta admits it "missed the mark" and is overhauling Meta AI's suggestion mechanism after a viral video showed the chatbot probing an Instagram user's personal family details. Company spokesperson Dina El-Kassaby confirmed to The Verge that the feature was never meant to surface that kind of prompt.

The video, posted last week by Instagram user Kalie Robins, showed Meta AI offering the prompt "Who is the child passenger?" beneath a clip of her singing with her daughter in a car. When she tapped it, the chatbot assembled information about her daughters from her earlier posts and from relatives' posts, then suggested follow-up questions about the children's ages and the family's location. Robins says the system also surfaced photos of her daughters that she had deleted years earlier.

According to El-Kassaby, the feature is designed to help people learn more about topics or posts that interest them, and responses are supposed to draw only on information the user already has access to. In this case, Meta AI appears to have pulled details from videos where Robins had named her children. The company says it has fixed the bug that triggered suggestions on personal subjects.

This is not the assistant's first stumble. In July Meta pulled a feature that allowed users to create deepfakes.