Meta's AI agent Muse risks privacy by exposing shared conversations

Meta's new general-purpose AI agent, Muse, has been found to expose sensitive user information when conversations are shared via web links, according to security researchers. Muse, which Meta CEO Mark Zuckerberg described last month as 'built from the ground up for privacy and security,' allows users to share content through unique URLs. However, a design flaw in the URL generation process made these links easily discoverable. Researchers from Claroc were able to enumerate 6,800 links created by Muse users for sharing conversations. Upon examining a sample of these links, they uncovered information that was clearly not intended for public access, including personal and business details. The issue highlights a potential privacy risk despite the platform's stated focus on security.