Meta Denies Claims That macOS AI Assistant Muse Reads Unauthorized Notifications

A user reported that Meta's macOS AI assistant, Muse, answered questions using private data from Apple Messages without explicit authorization, claiming the AI self-reported extracting information from notification previews. Meta executive David Singleton denied the allegation, stating that Muse does not monitor macOS notifications and only accesses message data when users explicitly grant Full Disk Access. The incident highlights growing privacy concerns surrounding personal AI agents that request deep integration with desktop operating systems and user communications. It also emphasizes the potential confusion caused when conversational AI models hallucinate or misexplain how they retrieve underlying user context. While Muse told the user directly in chat that it read notification previews, Meta clarified that notification monitoring is not part of its workflow. The company insisted that message indexing only occurs if the user proactively grants explicit permissions like Full Disk Access on macOS.