Visa and OpenAI partner to let AI agents make purchases autonomously
Visa and OpenAI announced a strategic partnership on June 10 that allows AI agents to initiate and complete financial transactions autonomously, integrating Visa’s payment infrastructure directly into OpenAI’s products including ChatGPT. The deal represents a major step toward a world where AI agents shop, pay bills, and manage subscriptions on behalf of their users — but it also raises unresolved questions about liability when an agent gets it wrong (ZDNet; Bloomberg).
How it works
Users link their Visa cards to their OpenAI account and set permissions for their AI agent: spending limits, approved merchant categories, transaction frequency, and whether human approval is required before each purchase. When a user asks ChatGPT to “find the best running shoes under $100 and buy them,” the AI agent searches, selects a product, and completes the purchase using the user’s Visa credential — without anyone manually entering card details at checkout (Visa; AP News).
The transactions flow through Visa’s existing payment rails, but the “shopper” is the AI agent. The underlying infrastructure uses Visa Intelligent Commerce, a platform Visa began building in 2025 that includes AI-Ready Cards (tokenized credentials with scope-limited permissions), the Trusted Agent Protocol for cryptographic agent authentication, and Visa Payment Passkey for biometric user authorization (BusinessWire).
The trust question
ZDNet’s coverage, published June 13, framed the announcement with a pointed question: “but can you trust it?” The answer is not straightforward.
Visa’s Trusted Agent Protocol, announced in October 2025, lets merchants cryptographically verify that an incoming transaction request comes from a legitimate AI agent acting on a consumer’s explicit authorization. The protocol has 100-plus global partners in its ecosystem, and Visa applies its existing zero-liability fraud protection policy to agent-initiated transactions.
But the security questions go beyond fraud. A user who tells an AI agent to “buy running shoes under $100” and gets charged $300 for a pair that was misidentified faces a scenario that existing dispute infrastructure was not designed for. Chargebacks911 warned in May 2026 that AI agents are creating “a new era of dispute risk” and that the infrastructure for handling chargebacks has not kept pace with the capabilities being deployed (Quartz).
Legal experts at Hogan Lovells flagged in February 2026 that under current payment regulations, consumers must be reimbursed for unauthorized transactions — but it is unclear whether an AI-agent-initiated transaction that exceeds the user’s intent qualifies as “unauthorized” or falls into a legal gray area in which the user authorized the agent but not the specific purchase.
A race between payment networks
The Visa-OpenAI partnership is one of several competing frameworks for agentic commerce, all announced within weeks of each other. On the same day as Visa’s announcement, Mastercard launched Agent Pay for Machines with 30-plus partners, using a “Verifiable Intent” cryptographic record of who authorized what instructions. Stripe’s Agentic Commerce Protocol, co-developed with OpenAI, has powered ChatGPT’s “Instant Checkout” feature since September 2025. Google and Shopify announced a Universal Commerce Protocol in January 2026.
Visa’s advantage is scale: its network supports 4.8 billion credentials globally. The company said AI-driven traffic to retail sites surged 4,700% year-over-year as of October 2025, and the Visa-OpenAI partnership is designed to capture that wave as it moves from product search to actual purchasing.
What comes next
OpenAI’s head of partnerships and commerce, Marco Mahrus, said the integration is about building “the infrastructure for secure, transparent, and user-controlled agentic transactions.” The companies plan to extend the capability to enterprise use cases, where AI agents in OpenAI’s enterprise platform Atlas could manage supply orders, software subscriptions, and recurring B2B payments within company-defined budgets.
The EU AI Act’s transparency obligations for customer-facing AI take effect on August 2, 2026. That date may arrive before the agentic payment infrastructure has fully answered the questions its own partners are asking.
Sources: [ZDNet](https://www.zdnet.com/article/openai-and-visa-aim-to-secure-agentic-transactions-how-theyll-work/) (June 13, 2026); [Bloomberg](https://www.bloomberg.com/news/articles/2026-06-10/openai-visa-team-up-to-let-ai-agents-make-purchases-online) (June 10, 2026); [Visa](https://corporate.visa.com/en/sites/visa-perspectives/innovation/visa-openai-partnership.html) (June 10, 2026); [AP News](https://apnews.com/article/visa-chatgpt-openai-shopping-mastercard-d769dec86344cb4977c98789e8ec492f) (June 10, 2026); [Quartz](https://qz.com/visa-openai-chatgpt-ai-agents-payments-061026) (June 10, 2026); [TechRadar](https://www.techradar.com/pro/openai-signs-major-visa-deal-so-ai-agents-will-soon-be-able-to-make-purchases-and-payments-for-you) (June 10, 2026)