Things May Get Ugly: Meta’s New AI Muse Could Make the Internet More Annoying

Last week, my finger hovered over the download button for Meta’s new AI assistant, Muse. For a moment, I felt strangely uneasy about my own phone.

Muse is designed to help with everyday tasks by acting as an AI agent—software capable of navigating the online world and carrying out tasks with limited human supervision. It can use its own web browser to cancel a gym membership, negotiate with customer service representatives, order groceries, invite friends to parties and much more. Soon, it is expected to make phone calls as well.

It is among the first free, full-featured AI agents released by a major technology company. Millions of people downloaded Muse within its first few weeks, while OpenAI has also announced its own new agent. More are expected to follow.

That could mark the beginning of a new chapter for the internet. But what will that chapter look like?

Experts who study AI agents warn that the transition could make the internet considerably more frustrating. These systems could trigger a chain reaction of problems that puts pressure on some of the basic infrastructure underpinning the digital world.

The internet was largely designed for humans, who have limited time, patience and attention. AI agents have none of those constraints. Instead, they can operate continuously and at enormous scale.

That could mean humans competing with AI systems for concert tickets, restaurant reservations and appointments. Small businesses could be overwhelmed by automated messages and requests. Websites that depend heavily on advertising could struggle as AI agents increasingly bypass ads. Users could also encounter more CAPTCHA tests and other systems designed to distinguish humans from machines.

And there is another concern: people who hand over significant control to AI agents could expose themselves to new security and privacy risks.

Eventually, the internet may be redesigned to accommodate this new reality. That future could potentially be more efficient and convenient. But the transition period may be far messier.

“Imagine there are no rules of the road. And you release billions of cars—the AI agents—and you just let them drive through playgrounds, hitting kids. That’s where we are,” said Ramesh Raskar, an associate professor at the Massachusetts Institute of Technology who studies AI agents. “It’s a pivotal moment.”

Muse also raises questions about how much access users should give an AI system.

To function, the application can request access to highly sensitive information and tools, including Gmail, calendars and, on Mac computers, extensive control over the device.

That prospect made me hesitate before downloading it.

Security researchers have also raised concerns about vulnerabilities in the application. Patrick Wardle, co-founder of the US nonprofit security organisation Objective-See, which identified serious flaws in the Muse app, said he would personally avoid using it.

“I wouldn’t use it,” Wardle said. “Personally, I would tell you to uninstall it altogether.”

The promise of AI agents is enormous: they could handle tedious digital tasks, save users time and make online services easier to navigate. But giving software the ability to act independently on our behalf also creates risks that the internet was never originally designed to handle.

The technology is arriving faster than the infrastructure and rules needed to contain it. Until the internet catches up, users may have to navigate an increasingly unpredictable online world—one where humans and AI agents are competing for the same digital space.