Meta Denies Claims That Muse AI Tool Read Private Messages Without Permission

Meta has pushed back against claims by a columnist that its AI tool, Muse, read users' private messages without permission, explaining that the application requires explicit consent and system-level protections for use.

Emergence of the Claims

Inc. columnist Jason Aten published a report claiming that Meta's AI tool, Muse, read users' private messages without authorization. This report sparked controversy in the tech world shortly after.

Statement from Meta

Meta Head of Communications Andy Stone strongly opposed the claims in a statement on social media, stating that the company does not believe the product performed this action without consent.

Stone emphasized that the Messages integration in the Muse app developed for Mac is completely optional and that users must activate it themselves.

Permission and Security Details

David Singleton, head of Meta Superintelligence Labs, explained that for the system to operate, multi-application permissions along with built-in macOS system-level protections must be triggered, and that this situation cannot be bypassed even due to a bug.

Andy Stone stated that for the Muse app to read messages, users are required to enable both Full Disk Access and the Messages connector.

Dispute Between the Parties

Writer Jason Aten claimed that at the time of the incident, Full Disk Access was turned off and that the AI was synchronizing device notifications. Meta defended that this explanation was erroneous and rejected the claims.

Meanwhile, YouTuber Matt Robb reported that a separate issue regarding Facebook Marketplace allowed permissions to share an address, but later admitted it was his own mistake.