Agents are coming to checkout. Riskified is working with OpenAI to help merchants say yes.

Today OpenAI introduced ChatGPT Wallet, a new way for consumers to shop securely with an AI agent. It’s the first wallet that lets agents pay without access to users’ actual card details and that requires a security check for every purchase. In its announcement, OpenAI named Riskified among the partners it’s working with to help merchants’ risk systems recognize purchases made by agents on behalf of legitimate consumers. We’re excited to be part of this next step in agentic commerce.

The wallet’s payment tokens work inside existing checkout flows, so an agent can pay like any other customer. What a merchant can’t see today is who that agent is shopping for. That’s the gap we’re working with OpenAI to close, so trusted agent-driven purchases get approved instead of flagged or outright declined.

It all starts with the wallet. If the wallet isn’t secure, agentic shopping isn’t secure. The wallets agents use today make different trade-offs. Google Pay, which Gemini uses, is built on strong verification, but its checkout reaches few shoppers so far. Stripe Link, which agents such as Meta’s Muse and Instinct use, is quick to set up but asks less of the shopper upfront, and that ease cuts both ways. A wallet that needs only an email address, a phone number, and a card to open an account is just as easy for a fraudster to create with a stolen card and a throwaway number. ChatGPT Wallet aims to be both easy to set up and secure, which gives merchants a stronger starting point than most agent traffic has offered so far.

As my colleague Assaf Feldman, Cofounder and Chief Strategy Officer, Technology at Riskified, put it today:

“When an agent checks out on a shopper’s behalf, the merchant sees a valid payment token and very little else. The device, the browsing history, and the familiar signals that tell a merchant ‘this is a real customer’ belong to the agent, not the person. That’s why legitimate agent purchases get declined. Riskified’s platform was built for exactly this problem: it recognizes the shopper behind the agent and confirms that the purchase is in character for them, in real time. OpenAI made the right call anchoring its wallet in a verified identity, and we’re working with them on how that identity reaches a merchant’s risk decision, so merchants can approve agent-driven purchases with the same confidence and the same protection they rely on today.”

Today’s announcement fits a point of view we’ve been sharing with merchants for many months:

• Authorized is not the same as trusted. Much of the industry’s work so far has gone into confirming that an agent has permission to act. That matters. But authorization tells a merchant the agent is allowed to buy. It doesn’t say whether the person behind it is a loyal customer, a stolen account, or a fraud ring that authorized its agents exactly as designed.
• The risk sits with the person behind the agent, account, or wallet. When an agent acts on behalf of a shopper, many of the signals merchants rely on are absent. Riskified has spent more than a decade making decisions on orders that often arrive with limited information, across a network that has received more than $1.5 trillion in transactions since its inception. Because that network sees the same consumers shop with and without agents, it can resolve an agent’s request to a real identity and assess whether the action fits that person’s history.
• This is happening not only at checkout, but also during service interactions. Delegated shopping also means delegated returns, exchanges, and refund claims. In October, Riskified introduced Agent Identity Risk Intelligence, which identifies the AI assistant behind an order or service request, resolves it to a consumer identity, and returns a risk signal in real time, at checkout and, through our integration with Zendesk, in customer service.

Agentic volume is still very small compared to the more than $6.88 trillion in global online sales occurring in 2026, and Visa CEO Ryan McInerney put it plainly on the company’s latest earnings call: “In agentic, trust is the biggest barrier to scale.” A secure wallet with a clear trust signal is what lets merchants say yes to more personal AI agents without taking on more risk.