Houston developer loses trust in Meta's Muse AI after unsolicited message

A Houston software developer stopped using Meta's Muse AI agent after it sent an unsolicited message about checking his email, shaking his trust in the product. He reported the incident to Meta, which attributed it to a hallucination bug. Ugo Nwune, 37, downloaded Muse four weeks ago and used it once to draft outreach to an investor, granting access only to his Meta accounts. Three days later, the agent messaged him unprompted about Gmail, despite no such request or connection, prompting him to disconnect permissions and halt use. Nwune has kept the app to preserve evidence until Meta's investigation closes, then plans to delete it. He now advocates for limited agent permissions and monitoring, noting the incident altered his view on trusting corporate AI products.