Meta Muse Share Links Expose User Conversations, Claroc Researchers Find

Security researchers at the firm Claroc reported Monday that Muse, Meta's new general-purpose AI agent, can leak sensitive information when users share conversations via web links. Muse permits sharing of searches, conversations, images and similar content, a feature that typically generates a unique URL for loading a page hosting the material. Such links normally contain a long, hard-to-guess text sequence to prevent unintended access, and Meta CEO Mark Zuckerberg said last month that Muse was "built from the ground up for privacy and security." Claroc found that a central design decision in how the URLs are constructed made them easy to discover, allowing the researchers to enumerate 6,800 links that Muse users had created for sharing conversations. After opening a few hundred of those links, the researchers encountered information that was clearly not meant to be publicly available. The findings follow Ars Technica's September reporting on a serious zero-day vulnerability in the assistant, which the outlet described as extraordinarily privileged.