Muse, the new general-purpose AI agent from Meta, can expose sensitive information when users share conversations through Web links, researchers reported Monday. Like many online services, Muse lets users share searches, conversations, images, and similar content through a unique URL. Such URLs typically contain a long, hard-to-guess sequence of text to keep sensitive information away from unintended viewers. Meta CEO Mark Zuckerberg said last month that Muse was "built from the ground up for privacy and security," and it appears to take the same precautions. However, a design decision in how the links are constructed has made them easy to discover.
Researchers from security firm Claroc said they enumerated 6,800 links Muse users created to share conversations. After opening a few hundred, they found material clearly not meant to be public, including a candidate's notes for a confidential executive-search call with a frontier AI lab, a national marketing company's internal revenue review, a small-business owner's loan application, and a to-do list containing a bank fraud alert with a partial account number. Other exposed files included a real-estate investor's performance report with property addresses and exact return and mortgage figures, and a workplace's full-year staff vacation calendar.
"If you had asked us beforehand, we would have guessed that gathering these would take real effort," the researchers wrote. "It took one query."
The links are easy to find through passive DNS search services, which record historical domain resolutions once a share link is clicked. All Muse share links use the domain cf.metaaiusercontent.com, making them identifiable. Adding the conversation name to each URL, such as loan-application-draft-a4c9e2.cf.metaaiusercontent.com, further reveals what each link contains and makes the public discovery possible.
Source: Ars Technica · Summarized by HeadlinesBriefing