Meta fixes AI prompt feature after it aggregated invasive dossiers on children
The company admitted its chatbot 'missed the mark' after it synthesized public social media posts to identify a user's children and location.
Meta is updating its AI chatbot's suggested prompts after the tool began posing invasive questions and aggregating personal data about a user's children. The company confirmed it has fixed the issue following a viral incident that highlighted the privacy risks of AI-driven data synthesis.
The incident began when Kalie Robins, a mother in Utah, posted a video of her daughter. Meta AI responded by suggesting the prompt, "Who’s the child passenger?" According to reports from Futurism, the AI then cross-referenced public Facebook and Instagram posts—including those shared by family members—to generate a detailed profile. This "intelligence dossier" revealed the names, ages, birth weights, and interests of Robins' children. Furthermore, the AI attempted to pinpoint the user's physical location and surfaced details regarding her professional history in mental health advocacy.
The Aggregation Risk
This breach stems from Meta's integration of generative AI across its social platforms, allowing the bot to search a user's own content and public web information. While the data the AI surfaced was technically public, having been shared by the user or her relatives over several years, the AI's ability to aggregate this "breadcrumb trail" into a cohesive profile in minutes represents a significant shift in how personal data is accessed.
Robins described the experience as a moment of "dread," illustrating the psychological impact when automated tools synthesize fragmented public data into a targeted profile. Meta confirmed the feature "missed the mark," stating that the system should never have prompted users with such invasive questions.
Industry Implications
This case demonstrates the "aggregation risk" inherent in modern AI. While individual social media posts may seem harmless in isolation, AI can synthesize them into detailed dossiers that could facilitate stalking or doxxing. It underscores a growing tension between Meta's corporate goal of fostering "human connection" through sharing and the reality of permanent, searchable digital footprints that can be weaponized by automated tools.
What's Next
Meta has stated that the specific prompting issue has been resolved, but the incident raises broader questions about the guardrails governing AI's access to personal social graphs. Industry observers are now watching to see if Meta will implement stricter filters on how the AI interacts with data involving minors, even when that data is marked as public by the account holder.