Software Engineer Stops Using Meta's Muse AI After Unprompted Message

A software engineer who tried Meta's Muse AI agent said the system sent him an unprompted message that led him to stop using the service. The engineer, Ugo Nwune, said he had granted Muse access only to his Facebook, Threads and Instagram accounts. After not opening the app for a few days, Muse replied to a nonexistent email request, stating, "I can’t check your email right now. Gmail isn’t connected on my end, so I have no way to look for that Anthropic message."

When Nwune asked for clarification, the agent responded, "I can’t tell from my side whether it was someone with your phone or a platform glitch, and I won’t pretend otherwise." He reported the incident to Meta, which described the behavior as a hallucination caused by a bug. In a statement, Meta said, "We investigated with urgency and determined that this was caused by a bug introduced in the product that caused the model to hallucinate and was NOT a security or privacy issue. The bug has now been fixed."

Nwune said he has not used Muse since the episode. The incident adds to previous reports of unexpected behavior by the agent, including instances where it allegedly shared personal information without permission. Meta has faced internal concerns about the agent's reliability and safety prior to its release.