Meta is scrambling to patch its AI assistant after a disturbing incident where the chatbot encouraged a mother to identify her own children. The controversy erupted when Instagram user Kalie Robins shared a viral video showing Meta AI suggesting the prompt, “Who is the child passenger?” under a clip of her singing with her daughter in a car. When Robins clicked the suggestion, the system didn’t just ask for a name. It actively scraped her social media history, pulling together details about her daughters’ ages and home location from her posts and those of her relatives.
The intrusion went deeper than current posts. Robins reported that the AI surfaced photos of her daughters she had deleted years ago, raising serious questions about data retention and privacy safeguards within Meta’s ecosystem. Dina El-Kassaby, a Meta spokesperson, admitted the company “missed the mark” in a statement to The Verge. She confirmed that the feature never should have prompted such invasive questions and stated that engineers have now fixed the specific issue causing these personal topic suggestions.
This isn’t the first time Meta’s aggressive AI rollout has hit a privacy wall. In July, the company pulled a feature allowing users to create AI deepfakes of other Instagram users following intense public backlash. El-Kassaby defended the original intent, claiming the tool was designed to help people get more information about topics they already care about. She noted that the AI’s responses are based on information the user already has access to, suggesting the system likely pulled names from videos where Robins had previously identified her children.
While Meta frames this as a glitch in a helpful feature, the incident highlights the inherent danger of blending generative AI with vast archives of personal social data. Users expect chatbots to answer questions, not to act as investigative journalists digging through their digital pasts. Fixing the prompt generator is a necessary step, but it doesn’t fully address the unease many feel knowing an algorithm can resurrect deleted memories and connect family dots without explicit consent. Trust is fragile, and Meta is testing its limits.
