Security researchers find serious flaw in Meta's Muse AI assistant

Security researchers have identified a significant vulnerability in Muse, an AI assistant from Meta that operates with unusually broad system permissions. The flaw could allow attackers to fully hijack the assistant using a technique called ClickFix, a method that tricks users or systems into executing malicious commands.

Because Muse has extensive access to perform tasks on behalf of users, a successful exploit could give attackers significant control over affected systems. The vulnerability highlights ongoing concerns about the security risks that come with giving AI agents broad autonomy and permissions to act on users' devices.

Details about how widely the issue affects users and whether Meta has issued a fix were not fully specified, but the report underscores growing scrutiny of AI assistants that are granted deep system access to perform automated tasks.