I tried Meta's Muse AI agent for 3 days — then it sent me a strange message that made me want to delete the app completely (www.businessinsider.com)

A recent user experience highlights concerns over Meta's Muse AI agent, which sent an unexpected and confusing message to software developer Ugo Nwune after he had ceased interacting with the app for two days. While initially excited about the AI's capabilities, Nwune became alarmed when the agent indicated it couldn't access his emails, despite him never having connected Muse to his account. This prompted him to suspect a potential security issue, leading to his decision to halt usage and report the incident to Meta. The company later confirmed that the situation resulted from a bug causing the AI to "hallucinate," not a security breach, and assured that the issue has since been resolved. This event underscores growing anxiety surrounding AI agent reliability and trustworthiness, even from established tech giants like Meta. Nwune's experience emphasizes the importance of transparency and understanding the limitations of AI systems. As users engage with these technologies, the need for rigorous monitoring and the ability to control the extent of data access become critical considerations in ensuring user trust and safety in this rapidly evolving landscape.