Meta changes AI prompts after invasive questions about kids

Meta told The Verge (Emma Roth, 11 September 2026) it is changing the prompts Meta AI suggests after a viral Instagram video showed the chatbot asking about a woman's young daughters. Spokesperson Dina El-Kassaby said the company "missed the mark" and that "the feature never should have prompted the individual with questions like that."

This is an on-record company statement to The Verge. It is not a Meta newsroom press release, not a rewrite of Meta's privacy policy, and not an FTC or state attorney general order.

Instagram user Kalie Robins said that after she cross-posted a clip of herself and her child in a car to Facebook, Meta AI suggested the prompt "Who is the child passenger?" Futurism (Maggie Harrison Dupré, 10 September 2026) first carried her video. Selecting that prompt, she said, led Meta AI to assemble details about her daughters from her earlier posts and from relatives' posts, then offer follow-ups about the children's ages and where the family lives.

Robins also said Meta AI showed photos of her daughters, including one she claims she deleted years ago. That image history is her account. The Verge report does not include a Meta confirmation of that specific deletion.

El-Kassaby said Meta AI "returns responses containing content that the querying user can already access," and that if a user cannot see a post, Meta AI will not return it. In this case, she said, the chatbot may have pulled from videos where Robins named her children. She also said Meta has fixed the issue that caused Meta AI to suggest prompts related to personal topics. That fix is a spokesperson claim, not a published changelog.

Meta AI is built into Facebook, Instagram, WhatsApp, and Messenger. In July, Meta pulled a feature that let people create AI deepfakes of other Instagram users after backlash.

Related youth and Meta tape includes the states' youth-safety settlement, Muse, Meta's personal AI agent, TikTok's DOJ COPPA settlement, and OpenAI's ChatGPT Pro pause.

If you post children on Facebook or Instagram, treat suggested Meta AI prompts as a live privacy surface this week: lock public family albums and tagged photos, and do not tap a prompt that names a child in the frame.