Skip to content
Brief The AI-First Web ·

OpenAI's ChatGPT Wallet lets AI agents pay, so stores must spot real agents

The launch starts with US ChatGPT Pro users and shifts the trust problem from hiding the card to recognising the agent.

In brief
  • OpenAI's ChatGPT Wallet lets its dot agents pay with a token, not real card details. Each purchase needs a Face ID or fingerprint check.
  • The hard part moves to stores. Their firewalls and fraud checks must tell a real agent from a bot.

OpenAI said on Friday that its AI agents, called dots, can now pay for purchases through ChatGPT Wallet. American Express, Mastercard and Visa helped build it. A user saves a card once. Before each buy, the user checks the items, total and card, then confirms with Face ID or a fingerprint. CNBC reported that OpenAI takes no cut of sales for now.

Only ChatGPT Pro subscribers in the US, aged 18 or older, with eligible US-issued cards can use it at first. OpenAI said the payment network supplies a token for each purchase. It works like a card number made for one deal, so stores can keep their normal checkout. OpenAI also said fraud systems may still flag real agent orders. The sources give no date for wider release.

So the hard part is no longer hiding the card. It is telling a real agent from an unwanted bot. OpenAI's guide asks sites to recognise signed requests using Web Bot Auth. A signature lets a site confirm which agent is calling, instead of guessing from traffic. CNBC reported that Amazon blocks agent purchases. Leaders should ask three things. Does our firewall treat signed agent traffic differently? What does our risk provider do with a token-paid agent order? Who owns declined agent orders?

A WebPulse Brief: a short report of an important event, written by the WebPulse Newsroom with AI assistance and checked against the reporting below. When there is more to explain, we follow up with a full story. How we use AI.

Reporting: OpenAI, CNBC, Help Net Security.