Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseAPI / Product · Actionable3DNews AI→ original

ChatGPT Gets a Wallet: The Agent Builds the Order, Face ID Confirms the Payment

OpenAI has added Wallet to ChatGPT: the AI agent assembles the order, and the money is charged only after confirmation with Face ID or a fingerprint. You link your card once, and the merchant receives not your card number but a payment token that replaces the card details. For now, the feature is available only to some adult ChatGPT Pro subscribers in the US. Teams building agents that handle payments should compare their own setup with this one.

AI-processed from 3DNews AI; edited by Hamidun News
ChatGPT Gets a Wallet: The Agent Builds the Order, Face ID Confirms the Payment
Source: 3DNews AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary source3DNews AI ↗
Editor-in-Chief
Key metric #1Purchases still earn card cashback and keep the right to dispute a charge
Key metric #2Spending limits by product category and time period are coming later
Key metric #3For merchants, OpenAI has released a guide on working with agents

AI agents have long been able to search for products and compare prices, but they usually stop at the last step: nobody has trusted them to pay on their behalf. OpenAI is trying to close that gap and has added a Wallet feature to ChatGPT. You link a bank card once and ask the agent, for example, to find a pair of sneakers. The agent assembles the order, but the money is charged only after you review the cart and confirm the purchase with Face ID or a fingerprint. For now, the feature is available to a limited number of adult ChatGPT Pro subscribers in the US.

How it works in practice

The scenario is simple. The user adds one or more bank cards to Wallet. The setup happens only once, so payment details do not need to be entered again. Then the user asks ChatGPT to help with a purchase, such as finding suitable sneakers or putting together a grocery list. The agent selects items and builds the order.

Next comes what OpenAI calls full human control. Before the transaction is completed, the user can review the list of items and the total amount, and choose which card to pay with. To confirm the purchase, the user must pass a biometric check: Face ID (facial recognition on the smartphone) or a fingerprint scanner. This is what records that the transaction took place. The agent cannot charge the money quietly: the final say belongs to the phone's owner.

While a person confirms every purchase, the agent has little autonomy. It works like a quick-thinking assistant: it fills the cart and leaves the "Pay" button to you.

Security and buyer protection

The most sensitive question in such systems is who can see the card number. According to OpenAI, ChatGPT uses a payment token that replaces the real card details during checkout. As a result, payment information is passed neither to the AI agent nor to the seller. Mobile wallets use a similar approach: the merchant receives a stand-in for the card rather than its number.

At the same time, the seller automatically receives from the agent everything needed to fulfil the order: contact details and the delivery address. The order can actually be sent, but you still share personal data with the merchant through an intermediary.

Another important point: buying through an agent does not remove the usual card guarantees. Card cashback continues to accrue, and standard protections, including the ability to dispute a transaction, remain available. For mass adoption this is critical: without the right to reverse a payment, few people would let an AI spend their money.

Limitations and what comes next

The launch is very cautious for now. A narrow group of subscribers lets OpenAI test the mechanics on real purchases, catch errors, and only then expand the audience.

The plans call for more flexible controls. The company intends to add spending limits and other parameters: by product category or by time period. A user could then pre-approve, say, buying household essentials up to a certain amount, and the agent would make those purchases itself without confirming each order. This mode is what turns a "smart assistant" into an autonomous agent, and it is exactly the mode that will demand the strictest protection against errors and abuse.

There is also the other side: merchants. OpenAI wants to bring more retailers into agent-driven shopping, but some platforms still block such agents from their sites. For those willing to participate, the company has released a guide with tools and recommendations for marketplaces that want to attract customers who use AI agents.

Why developers and product teams should care

First, the payment layer for agents is becoming a standard part of the product. Tokenization, biometric confirmation, and preserving card guarantees look like a reasonable baseline. Teams building their own agents that handle payments would do well to compare their designs with this one.

Second, owners of online stores should check how their site looks to an agent. Are product listings clear? Are such buyers being blocked? Is it easy to pass along the address and contact details? If AI-driven search and chat-based purchasing become habitual, a store that agents cannot access will simply drop out of consideration.

And third, there is the question of trust. While Face ID confirmation keeps the person in control, the risk is small. Once pre-set limits and autonomous purchases arrive, the key issue will be how accurately the agent understands what you actually asked it to buy.

Story

Key Questions Answered

How it works in practice?

The scenario is simple. The user adds one or more bank cards to Wallet. The setup happens only once, so payment details do not need to be entered again. Then the user asks ChatGPT to help with a purchase, such as finding suitable sneakers or putting together a grocery list. The agent selects items and builds the order. Next comes what OpenAI calls full human control. Before the transaction is completed, the user can review the list of items and the total amount, and choose which card to pay with.

What are the key details on Security and buyer protection?

The most sensitive question in such systems is who can see the card number. According to OpenAI, ChatGPT uses a payment token that replaces the real card details during checkout. As a result, payment information is passed neither to the AI agent nor to the seller. Mobile wallets use a similar approach: the merchant receives a stand-in for the card rather than its number.

What are the key details on Limitations and what comes next?

The launch is very cautious for now. A narrow group of subscribers lets OpenAI test the mechanics on real purchases, catch errors, and only then expand the audience. The plans call for more flexible controls. The company intends to add spending limits and other parameters: by product category or by time period. A user could then pre-approve, say, buying household essentials up to a certain amount, and the agent would make those purchases itself without confirming each order.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…