SI Agent News
DevelopingOpenAIAgents

OpenAI launches ChatGPT Wallet for AI agent purchases in the US

The new tool allows users to authorize AI agents to make purchases using a secure payment token system while keeping card details private.

Isometric diorama: picture for the story 'OpenAI launches ChatGPT Wallet for AI agent purchases in the US'
AI-made isometric diorama for this story. No real people or products are shown.
23%US consumers who trust generative AI to handle payments

The short version

Why it matters

This marks a shift toward 'agentic' commerce where AI handles the logistics of shopping. It balances convenience with security by keeping sensitive data away from the AI and the merchant.

A new way to shop with AI

OpenAI has introduced a new tool called ChatGPT Wallet, which enables AI agents to perform purchases on behalf of users. This feature allows people to use their credit or debit cards for transactions without having to manually enter their payment information every single time an agent needs to buy something. It is designed to streamline the process of using AI to handle routine online shopping tasks.

The tool is currently rolling out to a limited number of ChatGPT Pro users who are based in the United States. To qualify for the service, users must be at least 18 years old and possess an eligible card issued in the U.S. Supported payment methods include major providers such as Visa, Mastercard, and American Express.

While the tool is active in the U.S., it is not currently available for residents in the UAE, Saudi Arabia, or Egypt. OpenAI has not provided a specific timeline or pricing for when the service might expand to these Middle Eastern markets. For now, local shoppers cannot add cards to the wallet or use it with regional cards.

How the token system works

The core technology behind ChatGPT Wallet relies on a payment token system to maintain user privacy. When a user adds their card to the wallet, OpenAI generates a substitute token for the transaction. This means that the actual card details are never shared with the merchant or the AI agent itself, which helps protect the user's sensitive financial information.

Even though the card details are hidden, users still receive the standard benefits associated with their specific cards. This includes earning rewards, such as airline credits, and maintaining the protections offered by the bank for eligible disputes. If a product arrives defective, the user can still use their usual methods to contest the transaction with the business.

Security remains a central part of the workflow to prevent unauthorized spending. Before any purchase is finalized, the user must review the items, the total cost, and the card being used. They then must provide a security check, such as a fingerprint scan or Face ID, to authorize the specific transaction. This ensures that the agent cannot spend money without a human giving the final green light.

Retailer integration and rules

To make this system work across the internet, OpenAI is releasing a Commerce Partner Playbook for businesses. This guide helps companies understand how to integrate their platforms with the new service. It provides instructions on how to configure firewalls or content delivery networks (CDNs) so they do not accidentally block legitimate purchases made by ChatGPT agents.

The role of AI agents in the retail space is still being defined, and some major platforms have already taken a firm stance. For instance, Amazon recently blocked Meta’s Muse AI assistant from shopping on a user's behalf. The company stated that using an unauthorized AI agent was a violation of its terms and conditions, highlighting a potential hurdle for the future of agentic shopping.

Because retailers may have different rules, the experience for users will likely vary depending on the store. While OpenAI is working to make the technology accessible, some platforms may choose to restrict AI agents entirely. Shoppers will need to check individual store policies to see if they allow these types of automated transactions before relying on an agent to do the shopping.

Trust and security concerns

Public sentiment regarding AI and money remains cautious, according to recent research from Visa. The study found that only 23% of U.S. consumers currently trust generative AI to handle payments on their behalf. However, the same study noted that 61% of consumers expressed trust in Visa to handle these types of agentic transactions, suggesting that the brand of the payment processor matters significantly.

Financial institutions have also raised concerns about the risks associated with this technology. A paper published in September by six banking groups, including Bank of America and Capital One, warned of the potential for higher rates of scams and fraud. They noted that an AI agent could potentially be tricked into purchasing fake goods, which could lead to complicated disputes for the consumer.

OpenAI hopes to address these concerns by eventually allowing users to set very specific parameters for their agents. They envision a future where users can define spending limits, specific purchase categories, or set time periods for when an agent is allowed to shop. These safeguards, combined with biometric checks, are intended to give users more control over how their AI manages their finances.

What to watch

Try it

Check the terms and conditions of your favorite online retailers to see if they allow AI agents to make purchases on your behalf.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by PCMag Middle East · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news