Meta Changes AI Prompt Suggestions After Chatbot Asked About a User’s Children

Meta says it is changing the prompts its AI chatbot suggests under posts after a viral video showed Meta AI digging for personal information about a woman’s young daughters. In a statement to The Verge, Meta spokesperson Dina El-Kassaby said the company “missed the mark,” adding that “the feature never should have prompted the individual with questions like that.” Futurism first reported the story.

Last week, Instagram user Kalie Robins posted a video describing what happened after she cross-posted a clip to Facebook of herself and her child singing in the car. Beneath it, Meta AI suggested the prompt “Who is the child passenger?”

When she selected it, Robins says, Meta AI assembled information about her daughters from her earlier posts and from posts by her relatives, then suggested more prompts about her children’s ages and where she lives with her family. She says it also showed photos of her daughters on the platform, including one she says she deleted years ago.

What Meta says the AI can see

El-Kassaby said the system only surfaces content the user can already view. “Meta AI returns responses containing content that the querying user can already access,” she said. “If the user can’t see a post, then Meta AI won’t return it.” She said Meta AI may have drawn on videos in which Robins named her children.

Meta says it has fixed the issue that caused the assistant to suggest prompts on personal topics. “We created this feature to help people get more information about topics or posts that interest them,” El-Kassaby said. “When someone clicks on a prompt and uses this feature, the response it generates is based on information the user already has access to.”

Meta AI is built into Facebook, Instagram, WhatsApp and Messenger, where it answers questions, writes posts and generates images. In July, Meta pulled a feature that let people create AI deepfakes of other Instagram users after a backlash. Separately, ZDNET has explained how users can check whether their Meta AI prompts are visible publicly.