Riskified Joins Fraud Prevention Partners in OpenAI’s Announced ChatGPT Wallet
Key Facts
OpenAI announced ChatGPT Wallet on November 13, 2026, naming Riskified among the fraud prevention partners helping systems identify legitimate purchases made by AI agents. The company said the initial rollout would be through dots for ChatGPT Pro subscribers in the United States. The announcement puts Riskified among a group of companies helping develop payment infrastructure within ChatGPT, but does not make it the sole fraud prevention partner. Nor does it mean the agent will complete purchases without user intervention at launch. OpenAI said the service will begin with limited availability before expanding later.
The wallet is part of partnerships that include the American Express, Mastercard and Visa networks, and payment providers Adyen, Checkout.com and Stripe. OpenAI says users review the products, total and selected card before each transaction, then approve it with a security check such as Face ID or a fingerprint. After approval, the payment network issues a token that replaces the actual card details. OpenAI added that the wallet does not share actual card details with the merchant or agent. These steps put user approval and security verification at the center of the current payment process.
The initial rollout is limited to ChatGPT Pro subscribers aged 18 or older in the United States, with eligible cards issued there. OpenAI said ChatGPT is used by 1.2 billion people weekly, and that limited availability will allow it to learn from real-world use and improve compatibility and reliability. In a survey conducted by Riskified in September 2026, 43% of respondents said they wanted to approve every purchase before it was made. By contrast, 29% said they would be willing to let the agent buy automatically under rules they set. Only 7% were comfortable with the idea of AI buying a product without their approval.
The survey results show a divide between some shoppers’ acceptance of limited delegation and the preference of most respondents covered by the survey to review every transaction or set rules in advance. OpenAI’s announced launch mechanism aligns with requiring separate approval for each transaction, while the company said broader delegation could come over time. Under its plan, users will eventually be able to set spending limits, product categories or time periods for the agent to follow, alongside additional safeguards. That remains a future direction announced by OpenAI, not part of the initial purchasing process. For now, the purchase decision remains with the user after reviewing its details.
Riskified announced its Agent Identity Risk Intelligence capabilities on October 8, 2026. They are designed to identify the AI assistant behind an order or customer service request, link it to a real consumer identity and issue a real-time risk signal. The company said the capabilities will enter a limited pilot for enterprise merchants in December 2026, with general availability expected in 2027. The initiative provides context for Riskified’s work on risks in agent-driven commerce, while OpenAI did not provide further operational details about its role in the partnership in its wallet announcement. OpenAI also named Forter, Sardine and Signifyd as fraud prevention partners. The announcement therefore includes more than one provider for this aspect of the service.
Riskified reported revenue of $115.2 million in the third quarter of 2026, up 41% year over year, compared with growth of 22% in the second quarter. The company attributed the acceleration in revenue to winning new merchants and increasing sales to existing customers. Those results preceded OpenAI’s announcement and therefore do not reflect any publicly reported financial impact from the partnership. Visa, for its part, said 54% of failed test runs in its first AI shopping test nevertheless ended with a purchase made by AI. Visa plans to expand its test to include payment and checkout, according to the company.