Meta’s Muse: Unsettling in Unexpected Ways
Meta’s Muse is shaping up to be a powerful AI assistant, but it’s also raising some eyebrows with its slightly unsettling capabilities. The new Mac app can tap into Messages, Calendar, and Notes, which adds to its functionality. However, it seems Muse struggles to accurately describe its own operations. Jason Aten, a contributing editor, shared on Threads some screenshots of his interaction with Muse, where the assistant posed questions about a conversation he was having in Messages.
The catch? Aten claims he never granted Muse access to his messages. When he inquired how it was aware of his message content, Muse responded, “I saw the notification previews, not your message history. I haven’t been reading your texts.” When pressed further about how it received those previews, Muse didn’t provide a clear answer, stating, “Honest answer: I can’t give you the exact plumbing. What I know is that the paired Mac app exposes notifications as one of its capabilities, and they arrive to me through the device sync.” Not exactly comforting.
David Singleton from Meta Superintelligence Labs jumped into the discussion to clarify. He outlined the permissions Muse requires to read messages, including the need for full disk access for the Mac app, emphasizing that these features are opt-in. The more intriguing part of his response came later in the thread, where he noted that Muse “does not watch notifications on your Mac, but rather syncs data from Messages only after the user has specifically enabled access.” He suggested that the real issue was Muse’s misunderstanding of its own features, leading to the incorrect explanation.
In Singleton’s words, “In the conversation with his Muse in Jason’s screenshots, when Muse said it synced ‘device notifications,’ it was confused about how to explain the feature and gave an incorrect explanation. That’s on us. We apologize for the incorrect response from Muse and we’re working to improve Muse’s understanding of its own internals so that it gives correct answers to questions about how it functions more consistently.” While this may not be the most reassuring news, it does shed light on the challenges of AI communication. After all, we know that chatbots often prefer to give answers that sound good rather than revealing their inner workings.
