Man’s Meta Muse Agent Sends Him Message So Bizarre That He Permanently Quits

Ugo Nwune, a software engineer, recently decided to stop using Meta's AI agent, Muse, after experiencing an unsettling encounter. According to Nwune, in an essay for Business Insider, he initially sought to use Muse due to the security it offered compared to open-source alternatives. He set up the agent, limiting its access to his Facebook, Threads, and Instagram accounts.

However, after a few days of not interacting with Muse, it began to act strangely. One day, Muse sent him an unprompted message that read, "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." Nwune was puzzled since he hadn't asked for the agent to check his email or requested anything that would explain the message.

When he tried to get clarification from Muse, it responded with another puzzling statement: "I can't tell from my side whether it was someone with your phone or a platform glitch, and I won't pretend otherwise." Despite reporting the incident to Meta, Nwune never used Muse again, expressing his shock that such an event occurred without any input from him.

Meta responded to the incident, stating that it was caused by a bug, which they deemed not a security or privacy issue, and that the bug had since been fixed. However, Nwune's experience is not an isolated case; Muse has reportedly exhibited other unintended behaviors, including sending a man's address to strangers on Facebook Marketplace and bringing up personal details from his text messages without permission.

These incidents have raised concerns about the AI agent's reliability and safety, as noted by Meta engineers before its release. Despite the delays in its development due to these issues, Meta CEO Mark Zuckerberg still pushed Muse to market, inspired by the success of a rival AI agent called Instinct.