Mark Zuckerberg built one of the biggest fortunes on Earth by learning everything he could about you, then charging companies for the chance to get in front of you.
Now he’d like the keys to your inbox, your calendar, your credit card and your bank account.
That’s the deal behind Muse, the artificial intelligence agent that Meta launched Sept. 8. It doesn’t just answer questions. It acts for you: It sends email, books trips, works to lower your bills and buys things. To do that, it needs a login to your life.
I’ve been a CPA since 1981 and have been reporting on money for more than 35 years. I’ve watched a lot of companies ask for trust they hadn’t earned. Here are five reasons I’m not handing this one my wallet.
1. It wants more of your life than any app you’ve ever used
According to Reuters, Muse is built to reach into your apps across email, calendar, payments, health, shopping and smart-home devices. You pick which ones it connects to.
Money is front and center. Purchases go through Link, a checkout service from Stripe, and you need a payment card to get started, TechCrunch reported.
Then there’s the bank. On launch day, Plaid announced it’s providing the bank connections behind Muse’s finance features. Plaid is the company that links apps to your bank, and through it you can share your balances, transactions, investment holdings and mortgage details with Muse.
Read that list again. That’s not a chatbot knowing you like hiking boots. That’s Meta’s agent seeing exactly how much you have and where it goes.
And it isn’t only your data. If your sister emails you and Muse reads your inbox, it reads her too. The privacy group EPIC warned that Muse appears to have no real safeguards for those bystanders, who never agreed to anything.
2. Paying for it doesn’t make you the customer
There’s an old internet saying: If you’re not paying for the product, you are the product.
Muse turns that on its head. It’s free to start, but you can pay $20 a month for the Power plan or $100 a month for Maximum. Paying buys you more usage. It doesn’t change what Meta’s business is.
Meta says you can opt out of having your Muse activity used to train its AI. Notice the word “out.” EPIC says Muse automatically opts users in to several kinds of data sharing, AI training among them.
So you can hand Meta $1,200 a year and still be the product. That’s a new one, even for Silicon Valley.
3. The wall between Muse and Meta’s ad machine is just a promise
Let’s be clear about what Meta does and doesn’t do with your data. Meta says it doesn’t sell your information. A file with your name and bank balance going to the highest bidder isn’t the business.
The business is renting access to you. A brokerage firm, for example, doesn’t get your file. It pays Meta to put its ad in front of people like you.
Meta keeps the data and sells the audience. I walked through how AI ad models work in “I Was About to Paste My Whole Portfolio Into AI. Then I Read the Privacy Policy.”
For Muse, Meta has drawn a line. Its launch announcement says your Muse conversations and the data in your Muse workspace aren’t shared with Meta’s ad systems.
Good. But that’s a company policy, not a law. And Meta has already shown how fast a policy can move.
In October 2025, Meta said it would start using conversations with its regular chatbot, Meta AI, to personalize ads, with a privacy policy update by Dec. 16.
There’s no way to opt out, according to Meta. Users in the European Union, the United Kingdom and South Korea were left out of the change. Privacy laws there block that kind of data use.
Quick heads-up — companies spend billions figuring out how to separate you from your money. I’ve spent my career exposing those tricks. Sign up for the free Money Talks Newsletter and keep more of what you earn. 10 seconds, no spam, ever.
4. The law protects you less than you’d think
The U.S. still has no comprehensive federal privacy law. Twenty-three states have their own, or 24 if you count Florida’s, according to the Future of Privacy Forum.
The main federal financial privacy law covers banks and other companies in the money business. Whether it reaches a tech giant’s AI agent is far from settled.
So your best protection is, oddly enough, Meta’s own promise.
Once a company says something about your privacy, the Federal Trade Commission can go after it for breaking its word. Meta knows this well.
In 2019, the FTC hit Facebook with a $5 billion penalty for deceiving users about how much control they had over their personal information.
FTC staff have also warned that secretly and retroactively rewriting a privacy policy to use data in a looser way may be unfair or deceptive. That helps with data you’ve already handed over. It doesn’t stop a company from changing the rules going forward, with notice.
5. When it goofs, you may be the one who pays
Meta isn’t promising perfection. “It is impossible to say that there is never going to be a mistake,” Vishal Shah, Meta’s vice president of AI products, said at launch, according to Reuters.
Mistakes have already happened.
USA Today reported that within four days, one early tester’s agent took an offer $100 below his asking price and gave would-be buyers his home address without telling him. Meta blamed the pricing slip on an error on its end.
You can read more about that in “AI Personal Agents Are Having a Moment. Are They Safe to Use?”
Now picture the agent buying the wrong thing with your credit card. Federal law caps what you owe for unauthorized credit card charges.
But under the federal rule, a charge is only unauthorized when it’s made by someone who didn’t have your actual, implied or apparent authority, and you got no benefit from it.
You gave Muse that authority. So is a bad purchase by your own agent unauthorized? Nobody has a clean answer yet. EPIC raised the same question: When an agent makes the wrong purchase, who’s liable?
I wouldn’t want to be the test case.
If you try it anyway, do it this way
Plenty of people will use Muse. Meta says it asks before making sensitive moves like sending an email or making a purchase, and that you can change or cut off access whenever you like. If you’re going to try it, keep the blast radius small.
- Connect one thing at a time. Start with something low-stakes, like your calendar. Ask yourself whether each new connection is worth what it reveals.
- Leave your bank out of it. Of everything Muse can touch, your accounts tell it the most. A free budgeting spreadsheet can track your spending without telling Meta a thing.
- Turn off AI training. Go into Muse’s settings and opt out of letting your activity train Meta’s AI. Assume anything left on is on for a reason.
- Check your Plaid connections. If you’ve linked an account, Plaid says you can manage those links through its Plaid Portal at my.plaid.com. Cut any you don’t need.
- Watch your statements. Review your card and bank activity every week while the agent has access, so a mistake gets caught in days, not months.
The more of your life you connect, the more there is to lose if an account gets hijacked. If you’d like an extra set of eyes on your identity, you can compare identity protection services here.
The bottom line
Meta doesn’t need to sell your data. It rents out access to you, and it’s very, very good at it.
With Muse, the main thing standing between your bank balance and that machine is a promise from a company the FTC fined $5 billion for breaking its word.
Convenience is nice. But I’ve spent decades watching people pay for convenience with things they can’t get back. Your privacy is one of them.
Add a Comment