Meta just patched a security vulnerability in Muse AI agent
What's the story
Meta has released a patch for its Muse macOS app, after the discovery of a critical security vulnerability. The zero-day exploit, discovered by security researcher Patrick Wardle, involved an undocumented setting in the Muse app. This allowed potential attackers with local access to redirect transcription processing from Meta's servers to their own endpoint, gaining access to user accounts.
Exploit details
Exploit allowed access to user accounts
The exploit was enabled by several design decisions, including cloud-based Muse dictation and unrestricted control of undocumented settings by any app.
Wardle's proof-of-concept attacks demonstrated the potential damage this vulnerability could cause, allowing him to take pictures and write malicious files without alerting users in many cases.
"We can manipulate the agent and leverage its privileges to do whatever we want," Wardle told Ars Technica.
Company statement
Patch issued within hours of report
In light of the vulnerability, Meta issued a patch within hours of the Ars Technica report.
The company downplayed the real-world security risks associated with this exploit, as it required local access to a user's device.
"This was a local privilege escalation attack, not a remote exploit," David Singleton from Meta Superintelligence Labs said on X.
"Nonetheless, we have issued a hotfix to the app."
Launch
Muse's success and challenges
Despite the security issue, Muse's launch has been a hit for Meta.
In its first 12 days, the estimated downloads of the Muse mobile app have outpaced ChatGPT's own 12-day debut in the US and Canada.
However, Amazon recently blocked Muse from accessing its e-commerce platform over claims that Meta never sought permission to do so.