SI Agent News
ReportedOpenAIAgents

OpenAI launches ChatGPT Wallet for secure agent payments

The new feature allows AI agents to complete purchases using temporary tokens, keeping your actual credit card details private from merchants and the AI.

18minimum age for US access

The short version

Why it matters

This moves AI from a tool that just suggests products to one that can actually execute transactions. It solves a major security hurdle by allowing automation without exposing sensitive financial data to third-party bots.

A new era for agentic commerce

OpenAI has begun rolling out ChatGPT Wallet, a new feature designed to let AI agents handle payments for users. This tool allows the AI to complete purchases without ever seeing or sharing the user's actual credit card number with the merchant or the agent itself. By creating a layer of separation, the system aims to make automated shopping safer for the general public.

The company built this infrastructure in collaboration with major payment networks, including Visa, Mastercard, and American Express. This partnership ensures that the wallet can interact with the existing global financial systems that most online retailers already use. It represents a shift toward what Luke Gebb of American Express calls the company's "first American Express Agentic Commerce Experiences integration."

Currently, the rollout is not available to everyone. Only ChatGPT Pro subscribers living in the United States who are at least 18 years old can use the service. These users must also hold eligible US-issued cards to participate in the initial launch of the wallet functionality.

How the token system works

The technology relies on a process involving stand-in card numbers, often referred to as tokens. When a user wants to make a purchase, they first add their card to the wallet once. For every individual transaction, the system generates a unique token that serves as a substitute for the real card number, including its own security code and expiration date.

Before an agent can spend any money, the user must review the specific items, the total cost, and the card being used. The user then provides explicit authorization through a biometric check, such as a fingerprint or Face ID. Once approved, the payment network issues the token, and the final charge is applied to the real card behind the scenes.

This method ensures that the user keeps all the benefits of their original card. This includes earning rewards and maintaining the protections provided by the card issuer for eligible disputes. Because the merchant only ever sees the token, the risk of the primary card details being intercepted or misused by the agent is significantly reduced.

Challenges for online retailers

While the wallet makes it easier for users to buy things, it creates a new set of hurdles for the websites selling the goods. Merchants will have to do most of the adjusting to accommodate these automated checkouts. Some websites might accidentally block agent purchases because their security systems do not recognize the automated behavior as human activity.

This creates a specific problem for fraud prevention. Retailers that have tuned their defenses to stop malicious bots may find that they are now blocking legitimate purchases made by ChatGPT agents. These systems will need to learn how to distinguish between a wanted bot performing a task for a user and an unwanted bot trying to scrape data or perform a cyberattack.

To address this, OpenAI published a Commerce Partner Playbook to help businesses adapt. The guide asks websites to configure their firewalls or Content Delivery Networks (CDNs) to recognize Web Bot Auth. This is a specific method used to verify signed requests coming from bots to ensure they are authorized by a human user.

Partnerships and fraud defense

OpenAI is not tackling the security and fraud issues alone. They are working directly with several major fraud vendors to ensure that agent purchases are processed smoothly. These partners include Forter, Riskified, Sardine, and Signifyd, all of whom help companies manage the risks associated with online transactions.

The company is also collaborating with prominent payment providers like Stripe, Adyen, and Checkout.com. These entities provide the underlying pipes for moving money across the internet. By involving these established players, OpenAI hopes to ensure that the new wallet can work with existing checkout flows without requiring a total overhaul of the internet's payment infrastructure.

Even with these partnerships, some friction may remain. The risk engines that sit in front of checkout flows might still occasionally decline a transaction. If a risk engine flags a legitimate agent purchase as suspicious, the transaction will end in a decline or a cancellation, requiring the user to try the purchase again.

The path toward automation

The launch of ChatGPT Wallet marks a significant step toward a world where AI agents can act as personal assistants for physical and digital commerce. By solving the privacy problem of sharing card details, OpenAI is removing one of the biggest barriers to letting AI perform real-world tasks on behalf of humans.

For the user, this means a smoother experience where an agent can find, select, and buy products based on specific instructions. The human remains in the loop as the final approver, providing a balance between the convenience of automation and the security of personal oversight. This model of 'human-in-the-loop' commerce is likely to become the standard for AI interactions.

As more merchants adopt the Commerce Partner Playbook and update their firewalls, we can expect to see more seamless integrations. The goal is to create a system where the AI can navigate the complexities of online shopping while the user maintains complete control over their finances and personal data.

What to watch

Try it

If you are a ChatGPT Pro subscriber in the US, check your settings to see if the ChatGPT Wallet option is available for your account.

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 helpnetsecurity.com · 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