TECHNOLOGY · VERIFIED DEVELOPMENT
Meta’s Muse AI Assistant Accesses Messages via Notification Previews, Raising Privacy Concerns
WHY IT MATTERS
The incident highlights a privacy gap in Muse’s data handling, prompting users and regulators to scrutinize how AI assistants access personal communications.
What happened
Meta’s new Muse AI assistant, now available as a Mac app, can read data from Messages, Calendar, and Notes. In a recent test by journalist Jason Aten, Muse asked about a private conversation he was having in Messages, even though Aten had not granted the app permission to access his messages.
When questioned, Muse explained it had gleaned the content from notification previews. The assistant also admits it cannot describe itself.
The incident shows that Muse can infer message content without explicit permission, potentially exposing users to unintended data exposure. Meta must clarify how the assistant gathers information and whether it respects user privacy settings.
PRIMARY SOURCES
Meta’s Muse is creepy, but maybe not for the reasons you think
The Verge · Terrence O’Brien · Discovery only; Vox Media copyright terms apply