Links created to share conversations from Meta's new AI agent Muse could be found easily and exposed sensitive personal and business details, security researchers have said, according to Ars Technica.
Researchers at the security firm Claroc said they found about 6,800 shared Muse conversation links. Opening a few hundred revealed material clearly not meant to be public, including a confidential job-search call with an AI lab, an internal revenue review, a loan application, a bank fraud alert with a partial account number and a staff holiday calendar.
The links were discoverable through commercial passive DNS databases, which record domain lookups, because every one sits on the same Meta content domain and includes the conversation's title. The researchers said gathering them took a single query. Meta chief executive Mark Zuckerberg had said last month that Muse was designed with privacy and security in mind.
Source: Ars Technica
