Meta's new AI assistant, Muse, has drawn criticism after appearing to reference the contents of a user's Messages app despite the user saying he had not granted it access. Inc Magazine contributing editor Jason Aten shared screenshots on Threads showing Muse asking about a Messages conversation, then giving vague and inconsistent explanations when pressed on how it obtained that information, initially claiming it had only seen notification previews before admitting it could not explain the "exact plumbing." The episode is notable less for any actual privacy breach and more for highlighting that AI assistants can confidently misdescribe their own capabilities.

Meta Superintelligence Labs' David Singleton responded to the thread, explaining that Muse's Mac app requires various permissions, including full disk access, and that features such as message syncing are opt-in rather than automatic. He clarified that Muse does not monitor notifications but instead syncs Messages data only once a user has explicitly enabled that access, and acknowledged that Muse's own explanation of "device notifications" was confused and incorrect. Meta apologised for the error and said it is working to improve Muse's understanding of its own functioning so it gives more accurate answers about how it works.