Meta's Muse, the company's new general-purpose AI agent, is exposing sensitive user information through its conversation-sharing feature, researchers from security firm Claroc reported Monday. The flaw stems from a design decision in how Muse constructs the URLs used for sharing, making the links easy