Meta Denies Muse AI Agent Read User’s Private Messages Without Permission

Meta is pushing back against claims that its new AI assistant, Muse, accessed a journalist’s private messages without permission. The dispute began after Jason Aten, a columnist for Inc., reported that Muse suggested ideas based on his recent private conversations. He said Muse had synced a large part of his Messages database on his Mac, even though he believed he had not given the app permission to access his messages.

Meta’s response is clear: Muse cannot read Messages on a Mac unless two separate permissions are enabled. One is a macOS setting called Full Disk Access, and the other is the Messages connector inside the Muse app. Both must be turned on for Muse to access message content, according to Meta’s communications head Andy Stone.

David Singleton, an executive from Meta’s Superintelligence Labs, explained that these permissions involve multiple steps and system protections that cannot be bypassed, even if there was a software bug. He said the Messages integration is fully opt-in and requires explicit user approval.

However, Aten insists his Full Disk Access was switched off and that he never approved message access. He also mentioned that Muse told him it was only reading incoming notification banners, not full message history. Meta disputes this explanation, calling it incorrect and confusing.

This disagreement highlights a tricky issue about how AI assistants handle user data. Meta stresses that users stay in control of what Muse can access and can revoke permissions at any time. Still, the lack of independent verification means the exact cause of Aten’s experience remains unclear.

Meta recently launched Muse as a personal AI designed to help with daily tasks and long-term goals. While it promises strong privacy controls and user choice, incidents like these show that trust in such technology is still fragile, especially given Meta’s past data controversies.