One Hidden Meta Muse Setting Could Let Attackers Turn the AI Assistant Into a Backdoor

A security researcher has demonstrated how malware already present on a Mac can exploit Meta's Muse assistant, potentially turning it into a backdoor. The attack works by altering a hidden setting, allowing an attacker to intercept dictated prompts intended for the assistant. This is made possible by the broad access granted to the Muse app by its owner. The vulnerability can be exploited when a user uses the microphone to dictate a prompt, with the input being diverted to the attacker instead of Meta. The researcher released a proof-of-concept on September 21 to illustrate the issue. The flaw is related to a hidden setting in the Muse assistant.