Meta's AI Assistant Muse Gave Incorrect Information About Message Access

Serdar HocamAuthor & Editor

Meta's AI assistant Muse made confusing statements about how it accesses user messages; company officials stated that the AI misunderstood its own internal operations.

◉ 0 views
Meta’s Muse is creepy, but maybe not for the reasons you think

Meta's new AI assistant Muse raised security concerns by making surprising and erroneous statements regarding its access to user messages. Meta Superintelligence Labs explained that the assistant miscommunicated its own technical infrastructure and that the situation was merely a hallucination.

Muse and the Mac App

Meta's new AI assistant Muse is drawing attention with its new Mac app, which can access users' Messages, Calendar, and Notes applications.

Conversation on Threads

Inc Magazine contributing editor Jason Aten shared screenshots of his interaction with Muse on Threads and noticed that the assistant knew the contents of the messages.

Statement from Meta

David Singleton from Meta Superintelligence Labs clarified the details of the situation and explained the permission processes by commenting under the post.

AI Misunderstanding Its Own Internals

Officials acknowledged that Muse was mistaken and used incorrect phrasing while explaining its function, stating that the situation entirely stems from the assistant not knowing its own internal architecture.