From The Verge · Terrence O'Brien · · 1m 20s
Meta's Muse doesn't know how it works
Meta's new AI assistant can access your messages, but it can't explain how or why.
In brief
Meta's new AI assistant can access your messages, but it can't explain how or why. AI systems can function effectively while remaining fundamentally unable to explain their own operations, which is a bigger problem than any single privacy breach. Originally reported by The Verge.
Meta's new creepy AI assistant
Meta's Muse is an effective AI assistant for Mac that can access your Messages, Calendar, and Notes. But users are concerned about how much it can actually see.
Muse admits it's confused
When asked how it accessed a user's Messages without permission, Muse claimed it saw notification previews instead. But when pressed for details, it couldn't explain the mechanism.
“"Honest answer: I can't give you the exact plumbing."”
Meta's explanation: it's all opt-in
Meta's David Singleton clarified that Muse only syncs Messages data after users explicitly enable access and grant the Mac app full disk permissions. All features are opt-in.
Muse was just confused
Singleton admitted Muse gave an incorrect explanation of how it works. The real issue isn't that it's secretly reading your texts, but that the AI doesn't understand its own internals.
“"Muse was confused about how to explain the feature and gave an incorrect explanation."”
Chatbots tell you what you want
This reveals a deeper problem: AI systems routinely confabulate explanations rather than admit ignorance. Muse isn't uniquely broken, it's just following the pattern of all chatbots.
“"Chatbots won't tell you their secrets and will instead tell you what you want to hear."”
The one thing
AI systems can function effectively while remaining fundamentally unable to explain their own operations, which is a bigger problem than any single privacy breach.
