Should you trust new AI agents with your money and data?

DETROW: Stephan, then, walk us through whether and how using an AI agent could save people money.

STEPHAN BISAH, BYLINE: Yeah, so what these bots are really useful for when it comes to money is finding the low-hanging fruit options, the kind of stuff that's really easy to forget about or just really time consuming. Think, like, canceling subscriptions. That's how Scott Burka saved a lot of money. He's a Muse user in Nashville.

SCOTT BURKA: They were, like, things my kids had subscribed to on their iPads or, like, a chess account and things of that nature. And I said, no, I have not been using those. And it actually helped me cancel approximately $100 worth of expenses a month that were completely just under the radar.

BISAH: Another good example is car insurance. AI agents can be really helpful with that tedious and confusing work of comparing rates and doing it proactively every time your renewal is up, just letting you know without even asking you first.

DETROW: How much do they cost though?

BISAH: Dots from OpenAI comes in about $100 a month for a subscription. Grok Bot starts at 30. Meta's Muse, which again, seems to be the most popular one right now, has both a paid and free version. And the Muse users I spoke with said free seemed to be doing more than what they needed.

RUWITCH: Yeah, Scott, one of the currencies that you pay with for these things, though, is personal data. I mean, these apps only really work for you if you give them access to your accounts, email, calendars, Google Drive or contacts, so they can make a date with somebody, for instance. As with human assistants - right? - you're going to want - you know, if you want them to, like, go shopping for you, it's hard to do that if they don't know, first of all, where you are, what your brands are, where you like to shop. So the more data you give these things, the more useful they are as a tool, and, in turn, the more data the companies have on you.

DETROW: And that's where I've struggled because a lot of these, you know, personal finance, keeping-up-with-it stuff is stuff that I personally really struggle with, but I'm very hesitant...

RUWITCH: Right.

DETROW: ...To turn over my passwords to AI. So I'm curious, John, what is Meta, which makes Muse, told you about how they are handling user privacy?

RUWITCH: Yeah, Meta says Muse is highly secure. Each Muse agent is designed to not be linked to other agents. It's also limited in how and when it can go on the internet. Meta says people are always in control of how much access Muse has to their accounts, what permissions you give the app or whether it can access your passwords. And also, if a user wants Muse to check every time before doing something sensitive, like sending an email or buying something, it will. Critics say agents - you know, these agents can do tasks that humans can do, but they can also fail at them. Jiaxin Pei is an assistant professor of information science at the University of Texas, Austin, and he studies AI agents.

JIAXIN PEI: They're not as reliable as human beings. They make stupid mistakes all the time. They are now trained to interact with third parties very well.

RUWITCH: Yeah, so he says caution is warranted. At this point, users should be careful what tasks they delegate to agents.

DETROW: A lot of AI agents in the news this year - and the headlines have not always been positive, often not - a lot about them going rogue in one way or another. Is that a concern here?

BISAH: No, not really so much in these cases. These are not these souped-up AI agents companies were experimenting with, the ones that really push the edge of cybersecurity. These ones, like, are supposed to be much more limited.

DETROW: OK.

BISAH: Instead, many of the mistakes that we just heard about, these are just dumb shopping mistakes in many cases. For example, like, I recently got a new gas grill. I wanted to buy some inserts to make it act like a griddle, and Muse confidently gave me this recommendation that looked good, but it would have been a waste of money 'cause they're just the wrong size. So these agents, they can find you deals you'd otherwise miss, but it pays to double check their work.

DETROW: Look, a griddle on a grill is great, but let's talk about higher-level purchases. I mean, like, you're connecting your credit card and your bank account through Muse, right? Like, that could go wrong very quickly.

RUWITCH: Yeah, these transactions though are usually going through third party apps just like regular e-commerce, right? So it's not holding onto your credit card info. Muse, for instance, can peek into your bank account if you give it that permission, but it can't move money. It can just look inside, maybe give you some budgeting tips.

BISAH: But a lot of banks are coming out and warning that without proper guidelines, AI agents could raise risk of fraud and scam. They all say it is not clear who is liable when something goes wrong - the bank, the AI company or you.

DETROW: That is NPR personal finance reporter Stephan Bisaha and tech reporter John Ruwitch. Thanks so much.

RUWITCH: You bet.

BISAH: Thank you.

Copyright © 2026 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

Accuracy and availability of transcripts for NPR programming may vary. Transcript text may also be revised to correct errors or match updates to audio. Audio may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio record. Please note that NPR platforms and products also include some programming produced by third parties; NPR cannot guarantee transcripts for that content and some transcripts may be produced using automated or other tools.