Zero-day flaw lets attackers hijack Meta’s Muse via transcription endpoint abuse

A zero-day in Meta's Muse AI lets attackers hijack the assistant by manipulating its cloud-based transcription endpoint, enabling any locally run app or terminal command to obtain the Muse authentication token and take full control of the account; security researcher Patrick Wardle demonstrated PoCs that could write files or snap photos, highlighting risky design choices around cloud dictation and broad app privileges. Amazon has begun blocking Muse on its site, and questions about Muse’s security and privacy remain despite Meta’s public statements.