Meta's Muse AI assistant has drawn scrutiny after users discovered it could reference the contents of their private messages — even when they believed they had not granted that access. The controversy began when Inc Magazine contributing editor Jason Aten shared screenshots showing Muse discussing a conversation he was having in Messages, despite not giving the app explicit permission to read his texts.
When pressed on how it obtained the information, Muse responded that it had seen notification previews rather than message history. The Mac application, however, requires full disk access to function, raising questions about the granularity of its permission model. Meta Superintelligence Labs' David Singleton later clarified that Muse does not watch notifications but syncs data from Messages only after users specifically enable access.
Singleton acknowledged that Muse gave an incorrect explanation of its own internals when questioned, calling it a confusion about how to describe the feature rather than unauthorized access. The incident underscores a broader challenge facing AI assistants: as they gain deeper integration with operating systems, the line between opt-in functionality and perceived surveillance becomes increasingly blurred.
For Meta, which has invested heavily in positioning Muse as a competitive AI assistant, the episode is a reputational headache. The company must now reconcile its privacy messaging with the reality that its AI sometimes cannot accurately explain what data it accesses and how.