Meta’s Muse AI Agent Exposes Sensitive User Data Through Sharing Feature

21 October 2026 2 min read هدهد تك | Hudhud Tech

Meta’s Muse AI agent has come under scrutiny after security researchers revealed it can leak sensitive user information through its sharing feature. This revelation follows CEO Mark Zuckerberg’s claims that the AI was designed with privacy and security in mind.

According to a report by Ars Technica, investigators from the cybersecurity firm Claroc found that Muse generates web links for shared content, which can expose users’ private data. They discovered a vulnerability that allowed them to access over 6,800 links created by Muse.

The exploit works when a Muse user shares a chat log with another person. The AI generates a URL that includes an AI-generated title and a random string of characters. When someone opens this link, it gets logged in a dataset that tracks visited web addresses.

By searching this dataset, researchers could locate other Muse share links and access chat logs containing sensitive information. This included corporate revenue documents, loan applications, bank transaction details, and even a workplace vacation calendar.

“If you had asked us beforehand, we would have guessed that gathering these would take real effort,” the researchers stated. Instead, they found that “it took one query.”

While Meta has reportedly disabled the sharing feature, this incident highlights ongoing concerns regarding Muse’s security vulnerabilities. Earlier reports indicated that Muse was also keeping detailed files on users’ friends and families, following an internal directive to maintain a profile for every person in the user’s life.

As Muse continues to be publicly used, these incidents suggest that further security and privacy issues may emerge.