Visa processes AI-driven payments for OpenAI, but can you trust it?

Visa is stepping into the future of AI-driven payments by securing agentic transactions within OpenAI’s systems. But as the technology races ahead, the question remains: can consumers and businesses truly trust AI agents with their money?

The partnership, announced Wednesday, aims to bring agentic commerce into the mainstream by integrating Visa’s Trusted Agent Protocol with OpenAI interfaces like Atlas and ChatGPT Shopping. This allows developers and merchants to accept payments directly from AI agents, while Visa provides authorization and security layers. Transactions operate within user-defined guardrails, including spending limits, approval thresholds, and permission layers that keep the buyer in control even when an agent executes the work.

For consumers, this means letting AI handle shopping research and routine purchases, complete with pre-set personalizations. Merchants benefit from a more seamless buying experience across AI-powered surfaces where consumers increasingly spend their time. Visa’s announcement also applies to OpenAI’s expanding suite of AI-powered products, though the company declined to specify involvement with the forthcoming “superapp.”

The security landscape for agentic payments is still evolving. OpenAI has been developing its agentic commerce arm since launching Instant Checkout, which uses the Agentic Commerce Protocol developed with Stripe. This joins Google’s Agent Payments Protocol and other similar initiatives aiming to make agentic commerce appear more secure. On the same day, Mastercard launched Agent Pay for Machines to scale transactions between agents.

Yet skepticism remains high. AI agents have been known to go rogue, and many users are nervous about handing over payment controls. Visa emphasizes user control, transparency, and security, using tokenized credentials, real-time authorization, and fraud monitoring. Transactions operate within user-defined permissions, including spending limits, merchant categories, and approval requirements.

Experts, however, point to new risks. “Even with strong traditional security controls like tokenization and fraud monitoring, agentic payments introduce new risks that existing systems weren’t designed for,” said Geoff Cairns, principal analyst at Forrester. “The challenge shifts from authenticating users to governing whether agents act within intent and policy. For consumers and businesses, the main concerns are unauthorized or mistaken transactions, liability ambiguity, and fraud that scales faster than traditional dispute processes can respond.”

Beyond payment security, the broader AI shopping environment has trust issues. As the Guardian reported, AI shopping assistants can surface scam sites as legitimate retailers. While the payment process itself isn’t at fault, the ecosystem of AI-assisted shopping still struggles with trust holes before a buyer even reaches checkout.

So, are agentic payments worth it at this stage? “The convenience benefits don’t inherently increase risk, but they shift authentication from explicit user interaction to continuous, risk-based validation,” Cairns explained. “Delegated authorization and ‘on behalf of’ controls become central to trust. This is a technical area that is still evolving.”