Meta's AI chatbot was caught automatically prompting questions about a mother's children, pulling family data from deleted posts and relatives' accounts—and the company's answer was to call it a bug they've now patched.
This is the logical endpoint of a business model that treats every photo, every name, every family relationship as free fuel for machine learning systems nobody consented to feed. And the solution being offered to parents is to stop sharing pictures of their own kids in public spaces these companies built and profit from.
Kalie Robins, a Utah mother, posted a video of herself and her child singing in the car to Instagram. Meta AI, embedded across Facebook, Instagram, WhatsApp, and Messenger, automatically suggested the prompt: "Who is the child passenger?" When Robins followed the suggestions, the system assembled information about her daughters' ages and where the family lives. It displayed photos of her children—including one Robins says she had deleted years ago. It also pulled data from her mother's Facebook page, content Robins never posted herself.
"I thought I was being so careful and ultimately that didn't matter," Robins told Business Insider. "It didn't even get that info from me. It got it from other people, so it didn't matter how careful I was being in this case."
Meta's response: they "missed the mark." Spokesperson Andy Stone told Business Insider the feature "never should have prompted the individual with questions like that and we've fixed that issue." The Verge reported a similar statement from spokesperson Dina El-Kassaby, who noted that responses are "based on information the user already has access to."
That defense collapses on inspection. If the AI surfaced a photo Robins deleted, she no longer had access to it. The suggestion to identify a child wasn't requested by Robins—it was offered unprompted by Meta's system. The company built the tool, trained it on user data, deployed it across platforms with billions of users, and the "fix" came only after a viral video forced their hand.
This isn't Meta's first AI privacy failure either. In July, the company pulled a feature allowing users to create AI deepfakes of other Instagram users after public backlash, The Verge reported.
The fallout extends beyond one viral post. Business Insider reports that professional family photographers—nearly all mothers themselves—are now announcing they will no longer share children's faces on social media or their professional sites. The concern is AI-generated child sexual abuse material. Photos posted publicly could be used to create deepfakes of specific children or end up in training data for models that generate this content. Criminal cases involving AI-generated CSAM have already drawn national attention.
Business Insider framed this as a cultural conversation about sharing kids' photos; The Verge buried the broader implications and treated it as a single product defect Meta had patched. Neither pressed the structural issue: these companies harvest data at industrial scale, build AI tools on top of it, and when the tools do exactly what they were designed to do—surface personal information about users—the company calls it an accident and tells parents to post less.
The math for families is straightforward: share a family photo, and a system you didn't build, can't audit, and can't opt out of will mine it for data, suggest your children's identities to strangers, and potentially train future tools on their faces. Meta says it fixed one prompt. It hasn't fixed the machine that generated it.
