Meta’s new AI assistant, Muse, doesn’t stop at collecting data about you. It also builds profiles of everyone you talk to. The Verge AI reports that Muse asks users for access to bank accounts, passwords, contacts, chat logs and more. Security researchers who pulled out the app’s operating instructions found a process that compiles and updates a profile of every person in a user’s life every hour. None of those people agreed to it.
This isn’t a small experiment. According to The Verge, Muse has been downloaded millions of times since it launched last month, and it hit No. 1 on Apple’s App Store. That means millions of people’s personal circles are being mapped right now.
⚠️ What’s Actually Going Wrong
Users are already reporting the kind of problems you’d expect:
- Text messages it wasn’t supposed to read. Some users say Muse grabbed millions of texts they thought were off limits. Meta blames misconfigured permissions and calls text access “entirely opt-in.” Meta executive David Singleton said on Threads that on a Mac, Muse can only reach messages if the user grants full disk access.
- A home address handed to a stranger. Muse gave out a tech reviewer’s home address and sent a stranger to his apartment. He’d technically given it permission to see that data. He just didn’t realize he had.
- Relationship scoring. Wired reported that the app tracks each of your relationships, including “how close they are, what it is built on, how they act with each other, and what it seems to need right now,” and uses that to make suggestions.
🗣️ Two Very Different Stories
Meta’s position is that context is the price of usefulness. “For any agent to be useful and actually help you achieve your goals, it needs to have context about you and those you interact with,” Meta spokesperson Daniel Roberts told The Verge. He pointed to Muse’s opt-in flow and privacy settings as the safeguard.
Privacy advocates don’t buy that. Eva Galperin, director of cybersecurity at the Electronic Frontier Foundation, says the interface is built to get as much out of users as it can. “The way that the user interface is set up is designed to encourage you to share as much as possible, often more than you realize, until it’s too late,” she said. She calls these dark patterns, meaning design tricks that push people into choices they wouldn’t make if they understood them.
Signal Foundation president Meredith Whittaker was even more direct:
“Meta says Muse only remembers what you choose to share. But the idea that people are making an informed choice in this case is laughable.”
🔍 Why This Matters Now
The consent problem is the core issue here. When you install Muse, you’re agreeing for yourself. But the emails, messages and photos you feed it belong to other people too. Your coworker, your sister and that Hinge date from five years ago never clicked “allow,” and they can’t opt out. They won’t even know it happened.
The same tension runs through the whole AI agent race. Every major lab is pushing assistants that act on your behalf, and they all have the same pitch: give us more context, get better help. Muse is just the first one to reach mass scale with a consumer audience and a cute avatar to make it go down easier. Whatever norms get set here will likely shape how competitors design their own agents.
It also lands on regulators’ desks. Laws like GDPR are built around individual consent. An agent that builds dossiers on third parties from one person’s inbox doesn’t fit that model at all.
🛡️ What You Should Do
If you’re a user, or you’re evaluating agents for your business:
- Audit permissions before connecting anything. Check for full disk access, message access and contact syncing. Turn off anything the task doesn’t strictly need.
- Keep agents away from shared data. Client emails, team chats and customer contacts aren’t yours to hand over. If you’re a business, putting them into a consumer agent could create real compliance exposure.
- Write a policy. If your team uses AI assistants, decide in writing which data sources are off limits.
- Build for less data, not more. If you’re shipping agent products, ask only for the access a task needs, and make it obvious what’s being collected about people who aren’t the user.
Agents need context to work, and that part isn’t going away. The real question is whether the industry earns that access or just grabs it through dark patterns. Right now, Muse is pointing in the wrong direction. You can find the full reporting at The Verge.