--° Loading... Locating...
A smartphone showing the ChatGPT app

ChatGPT Can Now Shop With Your Credit Card. Here’s How the New Wallet Works and How to Protect Yourself

OpenAI's ChatGPT Wallet lets its AI agents buy things for you with your card, using one-time tokens and a security check on every purchase. It's limited to U.S. Pro users for now. Here's how it works, where it can't shop, and how to protect yourself.

Center

Key Points

  • ChatGPT Wallet launched Nov. 13 for U.S. ChatGPT Pro subscribers 18 and older.
  • OpenAI's "dots" agents can find and buy items; you approve each purchase with a security check.
  • It uses one-time payment tokens from Amex, Mastercard and Visa, so the agent never sees your card number.
  • United Airlines and Amazon block AI shopping agents.
  • OpenAI isn't taking a cut for now and hasn't detailed how agent mistakes are handled.
  • Use a credit card, not debit: federal law gives stronger dispute rights.
Listen to our news podcast

OpenAI now lets ChatGPT shop for you with your own credit card. The new ChatGPT Wallet, launched Nov. 13, lets OpenAI’s AI agents, which it calls “dots,” find products and complete purchases, CNBC reported. For now it’s only for ChatGPT Pro subscribers in the U.S. who are 18 or older. Here’s how it works, and what to think about before you let an AI near your card.

A smartphone showing the ChatGPT app
A smartphone running the ChatGPT app. Photo: Jernej Furman, CC BY 2.0, via Wikimedia Commons.

How does ChatGPT Wallet work?

You add a card once. Then you ask a dot to find something, say a pair of running shoes in your size under $120, and it goes shopping. Before anything is bought, you approve the purchase with a security check. OpenAI built it with American Express, Mastercard and Visa.

The agent never sees your real card number. The card networks give it a one-time “payment token,” a stand-in number, security code and expiration date good for that single purchase. OpenAI says you keep your card’s normal rewards and protections. It isn’t taking a cut of sales for now, though it says that could become a business later.

Advertisement article banner article banner

Is it safe to let an AI agent use your credit card?

The tokens are a real safeguard: if a retailer gets hacked, a one-time token is far less useful to a thief than your card number. The bigger risk isn’t theft. It’s the agent buying the wrong thing, the wrong size or from a seller you’d never pick. OpenAI hasn’t spelled out publicly how returns or agent mistakes are handled, and it calls this “early days” for AI shopping.

Some stores don’t want agents at all. United Airlines said last week it will block AI agents from booking tickets, and Amazon blocks shopping agents on its site. That means a dot may not be able to shop where you usually do.

What protections do you have if an AI buys the wrong thing?

Your card’s rules still apply, and the type of card matters. Federal law gives credit cards stronger protections than debit cards. Under the Fair Credit Billing Act, you can dispute a credit card charge for something you didn’t accept or that wasn’t delivered as agreed, and the card company has to investigate. A few practical rules:

  • Use a credit card, not a debit card. A debit dispute ties up your own cash while it’s sorted out.
  • Read the approval screen like a receipt: item, size, seller, shipping cost and total.
  • Start with small, low-stakes purchases until you trust how it behaves.
  • Keep the order confirmation emails. If something goes wrong, they’re your evidence for a dispute.
  • Check your statement weekly, the same way you would with any new subscription or app.

Who is ChatGPT Wallet for?

Right now, a small group of paying Pro subscribers, and the rollout is deliberately limited. “It’s going to take time for the ecosystem to build the right infrastructure,” said Chaya Nayak, who leads payments at OpenAI. With more than a billion weekly ChatGPT users, though, this won’t stay niche if it works. AI agents spending money is also one of the risks behind the lawsuit seeking to freeze advanced AI development.

The BeezLoop Take

Requiring approval on every purchase is the right call. An AI that can spend your money without asking is a liability, not a convenience, and OpenAI is smart to start slow.

But “early days” is doing a lot of work here. Before this opens to everyone, OpenAI should publish plain rules on who pays when the agent gets it wrong. Shoppers shouldn’t have to find out in a chargeback.

The open question: will stores like Amazon and United let AI agents in, or will AI shopping get walled off before it ever goes mainstream?

Sources: CNBC

How We Sourced This

Written by Kevin Nordi

Kevin Nordi is a freelance writer with five years of experience covering politics, sports, and the everyday moments that shape people's lives. He holds a Bachelor of Science in Multimedia…

More from this author →

BeezLoop News is an independent online news, discussion, opinion, and blog publication. Our articles combine reporting with editorial commentary and analysis. See our editorial standards for how we handle sourcing and corrections.

Leave a Reply

Your email address will not be published. Required fields are marked *

Start typing to search

🔔

Stay Updated!

Get instant notifications for breaking news and important stories. We'll keep you informed!

Don't miss a story

Get the day's clearest news explainers in your inbox.