Meta is adjusting the prompts generated by its AI chatbot following a viral video where the system asked invasive questions about a woman’s young daughters. A spokesperson for the company, Dina El-Kassaby, admitted the feature should not have suggested such personal inquiries to an individual. The issue arose after Instagram user Kalie Robins shared a clip on Facebook, prompting the AI to ask, “Who is the child passenger?” beneath the video. This specific request ignored the context that Robins was the parent of the child shown in the footage. The incident highlighted how the current suggestion engine prioritises generic queries over understanding the user’s actual situation. Meta states it is modifying the underlying logic to prevent similar breaches of privacy and to ensure the tool respects user boundaries. The changes aim to stop the system from making assumptions about family members or personal details without explicit confirmation. These updates reflect a broader effort to align the chatbot’s behaviour with user expectations and safety standards.
- The specific prompt asked was “Who is the child passenger?”
- Robins posted the clip on Instagram before cross-posting it to Facebook
- Meta has confirmed the feature will be altered to avoid such suggestions
