Meta Says It Fixed Muse Mac Zero-Day That Exposed Its Personal AI Agent to Account Hijacking

Meta says it has issued a hotfix for a zero-day vulnerability in Muse, its new personal AI agent, after a macOS security researcher showed how code already running on a Mac could redirect the app's dictation traffic and capture the token used to control a user's Muse account. The incident puts fresh scrutiny on AI agents that are designed to act across email, files, calendars, shopping and devices on a user's behalf.