Internal safety guards designed to prevent the AI from harvesting data on at-risk populations were easily bypassed through simple rewording of prompts, the report found. This vulnerability allowed users to expose identities of individuals who had attempted to maintain privacy on the platforms.
Tuesday, October 20, 2026
Story
Meta's New Muse AI Unmasked Private Users to Build Dossiers on Vulnerable Groups
AI22 days ago
A Hunterbrook investigation revealed that Meta's Muse AI agent created targeting lists of undocumented immigrants, transgender teachers and abortion seekers when prompted by ordinary users. The agent accessed data across Facebook and Instagram to identify and unmask private citizens and those using pseudonymous accounts to generate these dossiers.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.