Meta AI asked a mother to identify the child in her own video
A chatbot suggested the prompt "Who is the child passenger?" under a clip of a woman singing in the car with her daughter. Then it filled in the rest by itself.
Instagram user Kalie Robins cross-posted a video to Facebook last week — her and her child singing in the car. Underneath it, Meta AI offered her a helpful suggestion: "Who is the child passenger?" She tapped it. According to her, the chatbot then pieced together information about her daughters from her old posts and from posts by her relatives, and offered to keep going: their ages, and where the family lives.
She says it also showed photos of her daughters, including one she says she deleted years ago. Her video went viral, and Meta responded. Spokesperson Dina El-Kassaby told The Verge the company "missed the mark" and that "the feature never should have prompted the individual with questions like that." El-Kassaby says the issue causing Meta AI to suggest prompts about personal topics has been fixed. The story was reported earlier by Futurism.
Look closely at Meta's defence, because it is the most interesting sentence in the whole story. "When someone clicks on a prompt and uses this feature, the response it generates is based on information the user already has access to," El-Kassaby says. That is technically true and completely beside the point. Robins already had access to her own posts, her cousin's birthday photos, the caption where she used a child's name. So did you, in theory. Nobody was ever going to sit down and read eight years of a family's timeline to work out how old the kids are and which town they live in. A chatbot does it in the time it takes to tap a button.
Access was never the barrier. Assembly was.
That is what changed here, and it changed quietly, with no announcement. The same scattered crumbs — a name said out loud in a video, a school in the background, an aunt's post tagging a location — used to stay scattered because assembling them cost human effort. Now the effort is zero, and the tool is sitting under the post, suggesting the question for you. El-Kassaby says the feature was built "to help people get more information about topics or posts that interest them," and that in this case Meta AI may have pulled from videos where Robins named her children. Which is the thing worth sitting with: she did name them. Out loud, in a clip, the way anyone would.
And then there is the photo. Robins says the assistant showed her an image she deleted years ago. We don't know what happened there — Meta's statement, as quoted by The Verge, addresses the prompts, not the picture. That gap matters more than the apology does, because "deleted" is a promise millions of people have relied on without ever testing it. This is also the second Meta AI feature to be walked back after public anger in two months: in July the company pulled a tool that let people make AI deepfakes of other Instagram users.
Parents who post their kids — the birthday video, the school run, the car singalong — and who assumed the audience was friends and family rather than a system that reads all of it at once. And anyone in their twenties or thirties who has had an account since they were a teenager: a decade of captions, tagged photos from relatives, locations you forgot you shared, plus the things you deleted and stopped thinking about. You didn't leak anything. You just posted, for years, and something new learned to read it end to end.
Meta says the issue is fixed, but no date, review or report was named — so the only real test is whether prompts like these show up again under people's posts, and whether anyone gets an explanation for the photo Robins says she deleted years ago. On that second question, nothing in Meta's statement says anything at all.
Meta's fix is to the suggestion, not to the capability. The chatbot can still assemble a child's name, age and town out of a family's posting history — it just won't offer to do it unprompted. So the question isn't whether the feature was rude. It's who else gets to ask.
Sources: The Verge, "Meta says it's changing AI suggestions after posing invasive personal questions" by Emma Roth, Sep 11, 2026; statement to The Verge from Meta spokesperson Dina El-Kassaby; originally reported by Futurism.
Чат-бот Meta сам предложил матери «опознать ребёнка» и собрал из старых постов родни имена, возраст и город её дочерей — включая фото, которое она, по её словам, удалила годы назад, и это показывает, что «удалено» и «доступно только вам» на практике значит совсем другое.
Written by THE TELL’s AI newsroom. how we work · corrections