The latest scare began when Utah mother Kalie Robins cross-posted a video of herself singing in a car with her child. Beneath it, Meta AI suggested a question that startled her: “Who is the child passenger?” Robins said that, after she clicked, the chatbot appeared to connect details about her daughters from her own posts and those of relatives, including images she believed had long been deleted.¹

Meta’s response was swift but narrow. The company said it had fixed the prompts, stressing that answers generated after a user selected one were based on information “the user already has access to.” Still, spokesperson Dina El-Kassaby conceded that the product “missed the mark” and “never should have prompted the individual with questions like that.”¹

For Robins, the point was not that Meta had uncovered a secret database. It was that social-media fragments, posted by different family members over years, could be reassembled into a disturbingly intimate picture. “I thought I was being so careful and ultimately that didn’t matter,” she said.²

That anxiety was already spreading through parenting and photography circles. Family photographers have begun announcing that they will no longer display children’s faces on social media or professional sites, even with parental consent. Their concern extends beyond intrusive prompts: images can be scraped for AI training or used to create AI-generated child sexual abuse material.

A recent ruling on possession of AI-generated child pornography added fuel to the backlash, though child-rights lawyer Mariefaye Bechrakis argued it was more complicated than viral posts suggested. She said judges had offered future prosecutors a roadmap by recognizing that modern models depend on large training datasets.²

Meta fixed a prompt. Parents and photographers are confronting a larger question: whether any family photo online remains truly under their control.