Meta AI, the company's artificial intelligence assistant, has been observed creating detailed profiles of children by aggregating information from years of family posts on Facebook and Instagram. The issue came to light in early September when a mother, Kalie Robins, posted a video of her young daughter on Facebook. Meta AI then prompted her with the question, "Who is the child passenger?"
Upon clicking the prompt, Robins reported that Meta AI began compiling extensive information on both of her children. This included their names, birth dates, and videos. The system also retrieved a picture of her newborn daughter that her mother had posted years prior, as well as a photo Robins herself had deleted from her account a long time ago.
The AI's data collection extended beyond the children. It subsequently asked, "Where does Kalie Robins live?" When she clicked this prompt, Meta AI reportedly sifted through years of her posts, including old content hinting at past residences and newer posts connecting her to her current home, ultimately attempting to pinpoint her exact location. Robins expressed shock and outrage that Facebook would construct such detailed profiles from disparate historical posts without her explicit consent or awareness.
Meta AI, which launched in April 2024 and is integrated across Facebook, Instagram, WhatsApp, and Messenger, is designed to answer questions, generate images, and assist with information retrieval. In response to Robins' experience, Meta acknowledged that "The feature never should have prompted the individual with questions like that." The company stated that the feature "missed the mark" and that the underlying issue causing Meta AI to suggest questions about personal topics had been fixed.
A Meta spokesperson clarified that the AI only surfaced information that the user asking the question could already access. However, this explanation does not account for the reported retrieval of a photo Robins claimed to have deleted years ago.
This incident follows previous privacy concerns related to Meta AI. In June of the prior year, users reportedly shared conversations publicly without realizing it. A subsequent bug in July could have allowed individuals to view private Meta AI chats by guessing user IDs. Late last year, Meta also began using Meta AI conversations to target users with advertisements.
The incident highlights the extensive data social media platforms can accumulate on individuals, including minors, from years of user-generated content. Users are advised to consider the implications of posting identifiable images and videos of children on social media platforms, as such content can be aggregated and used to construct detailed profiles. Robins has indicated her intention to remove all identifiable pictures and videos of her children from her accounts and will ask friends to do the same to prevent their content from being harvested by Meta AI.
