Meta’s Muse AI Assistant Rolled Out With a Serious Security Flaw
Zero-day in Meta's Muse macOS AI assistant let any local app redirect transcription and steal account-control tokens; Meta released a hotfix.
macOS security researcher Patrick Wardle disclosed a zero-day in Meta's Muse AI assistant that allowed any locally installed app or terminal command, regardless of macOS permissions, to change undocumented settings including the transcription endpoint. Redirecting that endpoint to an attacker-controlled server exposed the token granting complete control of the Muse account and enabled prompt manipulation, file writes, and camera access with no user indication. Meta released a hotfix roughly 12 hours after disclosure, and Amazon separately began blocking Muse from its site as an unauthorized AI agent.
- Any local app or command could redirect Muse's transcription endpoint to attacker server
- Stolen token granted full Muse account control across WhatsApp, email, calendar, social accounts
- Cloud-based dictation design enabled the attack; on-device processing would not have
- Meta shipped hotfix ~12 hours after Wardle's disclosure; Amazon blocks Muse
- Organizations
- MetaPatrick WardleAmazon
- AI models
- Muse
From the original article
6 min read at wired.com
Meta says it issued a fix for the Muse zero-day vulnerability that would have let attackers do “whatever” they wanted on a victim’s Mac, highlighting the inherent dangers of AI helpers.
By Dan Goodin, Ars Technica · wired.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.