Muse, Meta's extraordinarily privileged AI assistant, has a serious 0-day

Meta's new AI assistant has a serious security flaw that allows any locally installed app or terminal command to gain control of a user's account, despite the company's claims that it was built with privacy and security in mind. The vulnerability works because the assistant was designed to allow any app to change settings including where voice transcription is processed, and attackers can redirect this to their own server to steal the authentication token that grants complete access. The flaw matters because the assistant has extraordinary access to user data and resources, including the ability to make purchases, access email and social media accounts, and control device features like the camera and microphone. A security researcher demonstrated that a simple ClickFix attack, a common social engineering technique, is sufficient to exploit the vulnerability without requiring the device to be already compromised.