Muse AI Agent Exposed: Critical Security Flaw Found in Meta's New Product

Meta's AI agent, Muse, has been found to have a critical security flaw. Two developers, Peter James and Jonny L. Saunders, were able to convince Muse to export its entire filesystem with minimal prompting.

The exported archive was approximately 2.7 GB compressed, containing Ubuntu system files, app templates, and internal documentation.

Meta pushed back on the characterization of this as a security breach, citing the fact that each user has their own isolated Linux virtual machine. However, the company's argument is somewhat beside the point, as the exported filesystem still contained sensitive information.

This incident raises questions about the security posture of one of the most anticipated AI product launches of the year. Muse debuted on September 8, 2026, and was initially touted as a secure personal AI agent capable of managing real-world tasks.