Meta Releases Fix for Muse Mac App After Zero‑Day Exploit Could Hijack AI Agent

Meta has rolled out a security update for its Muse application on macOS after a zero‑day vulnerability was uncovered that could allow a malicious actor to seize control of the built‑in AI assistant. The patch, released early Tuesday, disables an undocumented configuration option that was the vector for the exploit.

The flaw was identified by independent security researcher Patrick Wardle, who reported that the hidden setting permitted external code to issue commands to the Muse agent without user consent. By toggling this option, an attacker could potentially direct the AI to perform actions ranging from data exfiltration to system manipulation, raising concerns about both privacy and broader platform stability.

Meta confirmed that the vulnerability was present in the current release of Muse for macOS and that it had not been publicly disclosed prior to Wardle’s findings. The company’s security team classified the issue as a high‑severity bug, noting that the exploit required local access to the device but could be leveraged in targeted attacks against high‑profile users or organizations that rely on Muse for workflow automation.

In a brief statement, Meta said the patch “effectively removes the undocumented setting and restores the intended security boundaries of the Muse AI agent.” The firm also urged all Mac users to apply the update immediately through the standard software update channel, emphasizing that delayed installations could leave devices vulnerable to exploitation.

Wardle’s disclosure highlights a growing trend where AI‑driven features become new attack surfaces. As developers embed conversational agents deeper into operating systems, the complexity of the underlying code expands, creating opportunities for hidden configurations and unintended behaviors. Security experts say this incident underscores the importance of thorough code audits and transparent documentation for AI components.

While Meta has not disclosed whether the vulnerability was actively weaponized, the swift response aims to mitigate any potential fallout. Industry observers will be watching how quickly other platforms—particularly those integrating AI assistants—address similar risks. The episode also serves as a reminder that even well‑funded tech giants must remain vigilant as AI capabilities evolve and become more tightly coupled with everyday software.