Zero Day Vulnerability Discovered In Meta’s Muse AI Assistant For MacOS

Patrick Wardle found a zero day in Meta’s Muse AI assistant for macOS that would allow an attacker to manipulate where Muse sends the user’s audio recording, capture the user’s access token, and fully compromise the user’s account (including access to WhatsApp, emails, calendars, and user-authorized purchases). An attacker could perform these actions in a single line of code using a ClickFix style attack. Meta has since pushed out a hotfix (~12h after disclosure).

It’s a reminder of the risks associated with multi-functional agents. The more functionality you integrate into your agent, the greater the potential impact of a vulnerability. Unfortunately, the developers of Muse neglected to properly protect the communication channel between the user and Muse. As someone who relies on agents such as Claude Code every day, I always think about security, not convenience. While it only took Meta ~12h to release a hotfix, early adopters who integrated their personal accounts have already been compromised.