Researchers from security firm Claroc warned on Monday that Muse, Meta’s new general‑purpose AI assistant, may reveal sensitive information when users distribute conversations via web links. Like many online platforms, Muse lets people share searches, dialogues, images and similar content by generating a unique URL that anyone with the link can open. These URLs normally contain long, random strings to prevent accidental discovery.
According to the researchers, the design of Muse’s link‑generation system makes the URLs much easier to predict. By systematically scanning, they identified roughly 6,800 links that users had created for sharing their conversations. After opening a few hundred of those links, the team uncovered data that appeared to be unintentionally exposed, ranging from personal details to confidential business information.
Meta CEO Mark Zuckerberg has previously claimed that Muse was “built from the ground up for privacy and security.” The new findings suggest that, despite those assurances, the platform’s sharing mechanism may need additional safeguards to protect user privacy.