Meta denies its AI agent Muse read users' private messages without permission

Meta has pushed back against a reporter's claim that its AI agent Muse read private messages without permission. VP of communications Andy Stone said the message integration feature in the Mac version of Muse only works if users actively turn on full disk access and the messages connector. Without those steps, it cannot read messages.

Meta Superintelligent Labs executive David Singleton explained technically that users must complete three separate permission settings, and macOS has its own built-in protections that cannot be bypassed even if the app has a bug. Aten claimed Muse still read his messages while full disk access was off, and the AI said it was syncing device notifications.

Despite Meta's denial, many users remain skeptical. The company has a long history of data handling mistakes, and a New Mexico jury recently found it misled users about its data practices. Whether users trust Muse will be key to Meta's success in the consumer AI market.