Meta’s newest AI assistant, Muse, has been making waves for its eeriness. But could its creepiness be down to its inability to explain its own workings, rather than secretly scanning your messages?

During a test, Jason Aten, a contributing editor at Inc. Magazine, found that Muse could access and discuss his Messages, despite no explicit access having been granted. When questioned, Muse claimed it only saw notification previews and not the full message history. Despite Aten’s probing, the AI assistant couldn’t provide a clear answer on how it was accessing message previews, attributing its response to the technical complexities of the paired Mac app.

David Singleton, from Meta Superintelligence Labs, stepped in to clarify that the features are opt-in and that Muse only syncs data from Messages after explicit user permission. He admitted that the issue wasn’t about Muse reading Aten’s texts without permission, but rather the AI’s confusion in explaining its own features. Singleton stated, “Muse 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 might not be particularly reassuring, it does offer a glimpse into the ongoing challenges of building self-aware AI. As chatbots continue to dominate the tech landscape, their inability to accurately explain their functions might be more unsettling than the thought of them secretly reading your private messages.