ChatGPT Is Reaching Into Your Wallet: Watch Your Bot Closely

ChatGPT Is Reaching Into Your Wallet: Watch Your Bot Closely

I was two bites into dinner when my phone flashed a request: add a card to ChatGPT Wallet. I sat there, fork hovering, realizing the bot I use for drafts could soon ask to pay for things. You should feel that pause—I did, and it matters.

I’m going to walk you through what OpenAI announced, what actually changes for your money, and why most people are shrugging. I’ll point to the deals and the data so you can decide whether this is convenience or a slow creep. Think of me as the person who’s already read the fine print and will tell you the parts you’ll want to memorize.

My phone pinged with an onboarding prompt. How ChatGPT Wallet actually works

OpenAI says ChatGPT Wallet, built with American Express, Mastercard, and Visa, will let you add a card so your ChatGPT agent can prepare purchases without forcing you to re-type numbers every time.

That preparation still needs you: the agent proposes the item, you check the card and approve the purchase, and then you complete a biometric check like Face ID or a fingerprint scan. After approval, the payment network issues a token that moves the money without handing the merchant your raw card number.

How does ChatGPT Wallet work?

In practice, the flow mirrors existing mobile wallets such as Apple Wallet and Google Wallet: tokenization, preserved rewards and protections, and a biometric final step. What’s different is an added layer where an AI agent can act as your proxy if you set pre-approved rules—spending limits, allowed categories, time windows.

I read OpenAI’s privacy notice late at night. What OpenAI will log about your purchases

OpenAI’s privacy notice says Wallet will collect detailed purchase logs: merchants, items, prices, and shipping addresses. They promise not to sell that data or use it for targeted ads, but that still hands the company transactional patterns.

That sort of data is valuable. You’re not giving them the bare card number, but you are giving them a ledger of what you buy and where you send it.

Is ChatGPT Wallet safe?

Safety here is a mix: tokenization and biometrics are standard protections, and the card’s existing fraud protections and rewards remain intact. But safety isn’t just technical: it’s also behavioral and legal. How long records are kept, who at the company can access them, and what happens if the account is compromised are policies you should read before you trust autonomous spending.

Ride apps showed almost no AI orders last week. Why people aren’t using AI to buy for them

Uber reported only four rides and 500 Uber Eats orders placed by AI agents in the last week of October, out of about 320 million trips and deliveries. DoorDash said “maybe several hundred” agent orders total, and Airbnb called agent bookings “very non-material.”

Those numbers line up with a broader mistrust: an American Bankers Association survey found 77% of people are wary of giving an AI permission to make payments, and 61% said they’re unlikely to let an AI manage bank balances in the next 12 months.

Will AI agents place orders for me?

Technically, yes—especially if you’re a ChatGPT Pro subscriber (the tier starts at $100 (€92) per month in the U.S.). But adoption needs more than a technical path; it needs trust. Right now the numbers from Uber, DoorDash, and Airbnb show that users are mostly curious, not committed.

There’s a behavioral gap between convenience and comfort. People will hand a card to Apple Pay or Google Wallet because they already control their devices and habits. Handing pre-approved authority to an assistant is different: it asks you to think about delegating judgment, not just information.

OpenAI’s pitch is practical: faster checkout, fewer typos, and delegated chores. The catch is that the company will gain a precise map of consumer spending patterns. That’s less like handing a valet your keys and more like planting a camera in a room where you used to close the door.

Few companies have tried to make agents spend autonomously at scale because the demand isn’t there yet. That may change if the agents demonstrate reliability, tight limits, and transparent records of authorization. Or it may stay niche among power users who already pay for premium features.

If you’re deciding whether to connect your card, read the terms and privacy notice, test the approvals with strict limits, and watch for how the company handles data requests from third parties.

OpenAI is betting that convenience will win some users over, and partners like American Express, Mastercard, and Visa give the program instant credibility. But credibility is not the same as consent; real adoption will require gains in trust, not just functionality.

Would you let a script in your pocket spend your money while you sleep?