Meta says it is making changes to Meta AI's suggested prompts after a viral Instagram video showed it posing invasive questions about a woman's young daughters

Meta AI now avoids suggesting invasive questions about minors, a change directly addressing privacy concerns raised by a user.

Meta is implementing changes to Meta AI's suggested prompts following a viral Instagram video. In the video, Kalie Robins documented how Meta AI, integrated into Facebook and Instagram, generated invasive questions about her young daughters. The AI reportedly cross-referenced years of posts from Robins' and her family members' accounts, even when personal details like children's names were not explicitly shared by Robins. The AI provided detailed information including names, ages, birth weights, and even suggested school grades and local areas based on publicly available or shared content. Meta acknowledged the incident, stating the AI "missed the mark" and that the feature had been fixed, emphasizing that it was designed to help users find information they already had access to, but should not have prompted such intrusive questions.

This incident highlights a critical gap in AI's ability to interpret and leverage personal data from social media platforms, particularly concerning minors. While Meta claims the AI only used information users already had access to, its ability to synthesize and proactively suggest highly sensitive queries demonstrates a significant privacy risk. The company's swift acknowledgement and fix suggest an awareness of the potential for AI tools to overstep boundaries, especially as AI becomes more deeply integrated into user-facing applications.

The user's experience, and Meta's subsequent response, underscore the evolving tension between AI's data-parsing capabilities and public expectations for privacy, particularly for children. As AI systems become more sophisticated at aggregating disparate pieces of information, platforms must implement more robust safeguards to prevent the unintended creation of detailed personal dossiers. This event serves as a cautionary tale for how AI can inadvertently weaponize shared family memories and personal histories if not carefully governed.