Meta’s Muse AI Agent 0-Day Vulnerability Allows Attackers to Hijack the Tool and Inject Malware
A local zero-day in Meta's Muse agent lets existing malware redirect dictation, inject prompts, and steal credentials.
Patrick Wardle of Objective-See disclosed a macOS zero-day in Meta's Muse AI agent and released a proof of concept named not-a-mused. An unprivileged local process can change the undocumented endo_voyager_dictation_endpoint setting, redirecting dictation so an attacker can capture prompts, inject instructions, and obtain authentication data. The flaw does not remotely compromise a clean Mac; it amplifies malware already running as the user and could misuse Muse's access to files, mail, browsers, purchases, and linked devices. Meta had not publicly responded, while its bug bounty offers up to $300,000 for qualifying Muse flaws.
- Local code can redirect Muse via endo_voyager_dictation_endpoint without elevation.
- PoC not-a-mused captures prompts, injects instructions, and targets credentials.
- The flaw needs an existing user foothold and is not remote code execution.
- A hijacked agent could misuse files, mail, purchases, and linked devices.
- Meta's bounty offers up to $300,000; no public fix was reported.
- Vendors
- Meta
- Products
- Muse
- Organizations
- Objective-SeeMeta
- AI models
- Muse
From the original article
3 min read at cybersecuritynews.com
A zero-day vulnerability in Meta’s Muse AI agent for macOS could allow malware already running under a user account to hijack the assistant, intercept dictated prompts, inject malicious instructions, and steal authentication material. The flaw is especially concerning because a compromised agent could inherit the extensive permissions and connected-service access that users have entrusted to […]
By Guru Baran · cybersecuritynews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.