Meta’s Muse AI reportedly accesses Apple Messages without consent

Meta’s new personal AI agent, Muse, is facing serious allegations that it copied and uploaded data from Apple’s Messages app to its cloud servers, even after users explicitly denied it permission to do so. The controversy erupted shortly after the app’s September 8 launch. Tech journalist Jason Aten noticed something unsettling within 24 hours of installing Muse on his Mac. Despite denying the app access to Messages and disabling Full Disk Access, Muse started suggesting article ideas based on his private iMessage conversations. What Muse actually accessed When confronted, Muse initially claimed it had only accessed notification previews. But further investigation told a different story entirely. The app had synced over 187,000 rows of data from the user’s local Messages database. On macOS, accessing that database requires two layers of protection: explicit user approval and Full Disk Access permissions. Muse allegedly bypassed both. David Singleton, head of Meta Superintelligence Labs, acknowledged that the explanation initially provided to Aten was inaccurate. He accepted responsibility for the miscommunication while insisting that the access mechanisms were opt-in and not covert...