Meta Muse Flaw Could Have Exposed Users’ Emails and Files

An outside researcher reported a vulnerability in Meta’s Muse AI agent through the company’s bug bounty program. The flaw could have allowed an attacker to access a user’s dedicated virtual machine, which can contain emails and files. The reporting describes a potential exposure, not confirmation that anyone accessed or stole user data. 111

What the reported flaw could have exposed

Muse assigns users dedicated virtual machines—individual cloud-based environments. According to reporting based on an internal Meta incident report, the vulnerability could have provided an attacker access to one of those environments and to information stored there, including emails and files. 12

The public accounts do not explain the technical mechanism in enough detail to say how the vulnerability worked. They also do not establish that the reported flaw was exploited or that user data was taken. 111

Meta’s response and the severity rating

Meta was reported to be adding a clearer safety warning within Muse after the flaw was reported through its bug bounty program. The available reporting does not describe the exact technical fix. 13

Accounts of the flaw’s initial severity classification are inconsistent: one report summary calls it a high-severity SEV-2 incident, while another describes the initial tier as medium-to-high. The Reuters summary of the report does not give a classification, so the initial rating cannot be stated with confidence from the available sources. 1424

What Muse does—and why the risk matters

Muse is designed to carry out tasks for users, including shopping, booking travel, sending emails and making payments. 48 Because the agent operates with access to user-linked data and services, a vulnerability that could expose its dedicated environment raises a privacy concern beyond the agent’s ordinary task execution. 14

Muse was also reported to have topped U.S. app-download charts after launch, although the cited report provides no download total. 17