Meta’s new AI assistant, Muse, has a serious problem: a security researcher found a flaw that lets hackers take full control of a user’s computer through the assistant itself. The problem is a zero-day vulnerability, which means the flaw existed in the software when it shipped. It is not a small issue.
Jolly the Mascot
Muse comes with a mascot named Jolly, an adorable cartoon character that has become a talking point online. Some users have joked about wanting to have sex with the character. The appeal is obvious: Jolly is cute, friendly, and easy to interact with.
But the appeal is also the problem. The mascot creates a sense of trust that the software does not deserve.
What the Assistant Can Do
Muse is designed to perform a wide range of tasks. The company promises it can make purchases, generate images, create documents, and connect with other apps and services. It can also make phone calls, though those calls are made by humans, not the AI itself.
The assistant is meant to act as a personal helper, handling routine tasks for users. That convenience is the point. But convenience comes with risk, and Meta’s history with data has made users wary.
The Security Researcher
Patrick Wardle is a Mac security expert who discovered the flaw shortly after Muse was released. His findings were reported by Ars Technica.
Wardle’s discovery shows that the flaw was present from the start. That is the definition of a zero-day vulnerability.
How the Flaw Works
The flaw allows attackers to manipulate the assistant and use its privileges to do whatever they want on the victim’s computer. Instead of writing a complete malware program, an attacker can simply use the assistant itself as the tool.
“We can manipulate the agent and leverage its privileges to do whatever we want,” Wardle told Ars Technica last week.
That is a powerful statement. It means the assistant is not just a tool for everyday tasks. It is a vector for attack.
The False Sense of Security
Jolly’s cuteness is part of the problem. The mascot creates a false sense of security. Users see a friendly cartoon and assume everything is safe.
But the assistant has real privileges on a user’s computer, and those privileges can be abused. The flaw shows that assumption was wrong.
What Users Should Do
Users should treat any AI assistant with caution. The fact that Meta chose to use a mascot makes the situation worse, not better.
The advice is simple: stop giving Meta new opportunities to ruin your life. That is sound advice, even if it comes from a pop-culture website.
A Pattern Worth Watching
Meta has a history of data issues. The company has shown time and again that it cannot be trusted with sensitive information. This is just the latest example.
The Verdict on Muse
Muse is not delivering on its promises. The assistant cannot make purchases, generate images, create documents, or connect with apps and services. Those features exist on paper, but they are not working as advertised.
The phone calls are made by humans, which is a relief, but the security flaw is a much bigger concern. A hacker can take control of a user’s computer through the assistant, and that is a serious vulnerability.
The flaw is a zero-day vulnerability, which means it was present in the software when it shipped. That is a failure on Meta’s part.
The mascot is cute, but it is also a trap. Users should be wary of any assistant that asks for access to their computer.
Meta’s adorable AI is, predictably, a security nightmare.
See the video the story is built around at The A.V. Club.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.