Meta Denies Claims Its AI Agent Muse Reads Private Messages Without Permission

Meta executives Andy Stone and David Singleton publicly denied claims that the Meta Muse AI agent reads private Mac messages without user consent. Meta clarified that accessing message data requires explicit multi-layer authorization, including manually granting macOS Full Disk Access. User trust in privacy is critical for Meta as it launches autonomous AI agents capable of operating across personal apps. Given Meta's past privacy controversies, misunderstandings or AI hallucinations regarding background data collection could severely impact user adoption. Meta detailed that macOS security boundaries require users to manually enable Full Disk Access in System Settings, configure Messages app permissions, and restart Muse before any reading occurs. Meta attributed the journalist's concerns to an AI hallucination where Muse incorrectly explained its own system behaviors.