Meta rushes to fix AI chatbot privacy flaw - Bangkok Post
Meta has urgently updated its artificial intelligence (AI) prompt suggestion system after its chatbot prompted a mother to identify her young daughter and subsequently exposed detailed personal information, raising fresh privacy concerns among parents.
The incident surfaced after Kalie Robins, an Instagram user, uploaded a video featuring her daughter. Meta AI automatically generated a suggested prompt asking, "Who is the child passenger?"
When Robins interacted with the chatbot, it retrieved and displayed her children's names, ages, family details, location data and even photographs she believed had previously been deleted. Disturbed by the breach, she posted a video online questioning the platform's security and privacy measures.
Meta spokesperson Dina El-Kassaby acknowledged the error in a statement to The Verge, stating that the suggestion system had malfunctioned and diverged from company standards.
She confirmed that Meta had resolved the issue preventing the AI from suggesting prompts regarding personal data of individuals in videos, adding that the chatbot had only accessed information the user already had permission to view.
Despite the Big Tech company's assurances, the explanation has failed to quell anxiety among parents. Critics are concerned that the AI system actively processes and links historical posts across family members' accounts, including supposedly deleted content, to construct detailed personal profiles.Β
Robins noted that the AI appeared to synthesise details gathered from posts made by other family members to compile the comprehensive profile of her child.
