Meta has said it is changing the AI prompts suggested by its chatbot after a viral video showed it seeking personal information about a woman's young children. The move follows criticism from Instagram user Kalie Robins, who found that Meta AI generated an invasive suggestion, "Who is the child passenger?", beneath a video of her singing with her child, and then offered further prompts about her daughters' ages and family's location.

A Meta spokesperson, Dina El-Kassaby, said the company "missed the mark" and that the feature should never have prompted such questions, adding that the issue causing Meta AI to suggest prompts on personal topics has now been fixed. Robins said the AI also surfaced photos of her daughters, including one she believed she had deleted years earlier, having apparently pieced together details from her own posts and those of relatives. Meta said the AI's responses draw only on information users already have access to, and that in this case it likely picked up on videos in which Robins had named her children. The incident follows Meta's decision in July to withdraw a feature allowing users to create AI deepfakes of other Instagram users after a similar backlash.