Researchers found a 0-day letting attackers hijack Meta's highly privileged Muse AI agent through a simple ClickFix attack, raising enterprise risk questions.
Security researchers disclosed a serious zero-day vulnerability in Muse, Meta's new AI agent, that allows full hijacking through a basic ClickFix technique. The report notes Muse operates with extraordinary privileges across a user's accounts and tasks, making a successful exploit especially damaging.
The flaw surfaces just as Meta is racing to expand Muse into wearables, glasses, and email — widening the attack surface at the same moment the agent gains more autonomy and access.
Highly privileged agents are becoming attractive targets precisely because they're built to act on a user's behalf across accounts and purchases. For leadership deploying agentic AI, this is a concrete case for security review before granting agents broad permissions, not after.
The daily signal, curated. Get it in your inbox.
Subscribe on LinkedIn →