Meta Muse exposes users' personal and business secrets in shared conversations (arstechnica.com)

Meta's new AI agent, Muse, has been exposed as a security risk, revealing sensitive user information through easily discoverable shared conversation links. Researchers from Claroc reported that while Muse was designed with privacy in mind, its URL structure compromises confidentiality. The URLs, which include conversational titles along with unique identifiers, can be easily enumerated, allowing observers to uncover private discussions, internal documents, and personal data with a single query. This discovery raises significant concerns for the AI and machine learning community, highlighting vulnerabilities in data-sharing mechanisms that prioritize ease of access over security. Claroc researchers found sensitive information, including executive meeting notes, loan applications, and internal company reviews, by merely performing passive DNS searches. Such revelations could have serious implications for privacy and data protection, underscoring the need for robust security measures in AI-driven platforms that handle personal and business-critical information.