Meta adjusts AI prompts after invasive personal questions

Meta is revising its generative AI prompting features following reports that the assistant surfaced invasive questions regarding a user's minor children. The adjustment follows a viral video shared by Instagram user Kalie Robins, who demonstrated how Meta AI generated prompts about her daughter after she shared a car video. When clicked, the chatbot compiled details from her past posts and those of her relatives, while displaying photos that the user believed she had previously deleted.

Corporate response and safety measures

Meta spokesperson Dina El-Kassaby told The Verge that the company missed the mark and noted that the feature should never have surfaced such personal inquiries. According to El-Kassaby, the system is designed to retrieve only information that the querying user already has permission to view. The company claims it has now resolved the technical issue that led to the generation of these sensitive prompts.

The Scale Report perspective

This incident highlights a persistent tension between platform utility and user privacy in the age of integrated large language models. The Scale Report understands that while Meta argues its AI only aggregates existing visible content, the unexpected synthesis of scattered personal data across Facebook, Instagram, WhatsApp, and Messenger creates a psychological bridge that many users find alarming.

As AI assistants become more deeply embedded into social media feeds, the threshold for what constitutes helpful automation versus invasive data mining remains a moving target for developers. This is not the first time the company has faced scrutiny over AI behavior; it previously retracted a tool capable of creating deepfakes of users earlier this year. Managing user expectations regarding the 'memory' of these models will remain a central challenge for the engineering teams.