# Visa, Mastercard and Stripe are turning AI agents into payment actors Source: TechNewsList (https://technewslist.com) Canonical URL: https://technewslist.com/en/article/visa-mastercard-stripe-ai-agent-payments-2026-04-29 Section: Fintech (https://technewslist.com/en/fintech) Author: TechNewsList Language: en Published: 2026-04-29T19:47:00.087+00:00 Updated: 2026-04-29T19:47:00.254699+00:00 > Agentic commerce is moving from demos to payment plumbing as Visa, Mastercard and Stripe define how AI systems can safely go from recommendation to purchase. ## TL;DR - Visa says it and partners have completed secure AI transactions as agentic commerce moves toward 2026 adoption. - Mastercard Agent Pay frames AI agents as tokenized payment actors inside card-network trust rails. - Stripe documents machine payments for automated systems and AI agents. - The key problem is delegated authorization: agents need scoped payment power, not unrestricted card access. - Agentic commerce could reshape merchant discovery, fraud scoring and checkout infrastructure. ## Key points - Visa is positioning secure AI transactions as a foundation for mainstream agentic payments in 2026. - Mastercard Agent Pay focuses on tokenized credentials for AI-powered commerce. - Stripe machine payments documentation gives developers primitives for automated and agentic payments. - Agent payments require consent, policy limits, identity, audit logs and revocation. - Payment networks are trying to keep card rails central as AI changes the shopping interface. - Merchant data quality may become more important as agents choose products on users? behalf. - Liability and dispute handling remain major unresolved issues. # Visa, Mastercard and Stripe are turning AI agents into payment actors ## What happened Agentic commerce is becoming a real fintech category. Visa says it and partners have completed secure AI transactions, framing 2026 as the year agentic payments move closer to mainstream adoption. Mastercard has already introduced Agent Pay, a program for tokenized, trusted AI-agent transactions. Stripe, meanwhile, now documents machine payments for automated systems and AI agents. The shared idea is simple but technically heavy: AI agents should be able to discover products, compare options and complete purchases on behalf of users, but only inside payment rails that preserve consent, identity, limits and dispute protections. This is not just a checkout redesign. It is the beginning of a new payment actor. Until now, payment networks were built around humans, merchants, cards, wallets, processors and banks. AI agents add another layer: software that can initiate commercial intent before a person touches a button. ## Why it matters If AI agents become shopping, booking and procurement interfaces, payment networks need to know whether a transaction is truly authorized. A normal chatbot recommendation is low risk. A chatbot that can spend money is a financial actor. That creates several hard questions. How does a merchant know the agent is acting for a real user? How does a network enforce spending limits? How does the user approve categories, merchants, amounts or recurring actions? How do banks score fraud when the buyer is an AI workflow rather than a browser session? Visa and Mastercard are valuable here because they already operate trust systems at global scale. Their entry suggests agentic payments will not be solved only by app developers. It will be solved through tokenization, identity, network rules, liability models and merchant acceptance. Stripe's documentation adds the developer angle. If AI agents are going to purchase API credits, book travel, buy inventory or pay invoices, builders need primitives that feel programmable without becoming dangerous. ## Technical details The main technical concept is delegated authorization. A user should not hand an AI agent unrestricted payment power. Instead, the user grants a scoped permission: spend up to a certain amount, at certain merchant types, for a certain task, under a policy that can be revoked. Tokenization is likely central. Mastercard's Agent Pay framing points toward tokenized credentials that can represent an agentic transaction without exposing a raw card number. Visa's announcement similarly emphasizes secure transactions and partner work around moving from product discovery to purchase. The second layer is identity. Merchants and networks need to distinguish a human checkout, a bot attack and a legitimate AI agent. That may require agent identity, wallet attestations, device signals, merchant-side metadata and network-level risk scoring. The third layer is auditability. If an agent buys the wrong item, the user needs a record of what the agent saw, what instruction it followed, what policy allowed the purchase and where the authorization happened. Without that, disputes become messy fast. ## Market / industry impact Agentic payments could reshape online commerce because the interface moves upstream. Today, merchants compete for human clicks. In an agentic world, merchants may also compete to be chosen by AI shopping assistants. Product data quality, availability feeds, return policies, delivery promises and machine-readable trust signals become more important. For payment networks, this is a chance to stay central as AI changes the checkout layer. If Visa and Mastercard define trusted agent credentials early, they can make agentic commerce look like an extension of existing card rails rather than a threat to them. For fintech startups, the opportunity is around orchestration: policy wallets, agent spend controls, procurement agents, merchant-agent APIs, fraud analytics and consent dashboards. The winners will not simply let agents pay. They will make agent spending understandable and reversible. ## What to watch next Watch for merchant pilots. Agentic payments only matter if real retailers, travel platforms, marketplaces and software vendors accept them. Second, watch liability language. The most important details may not be in demos, but in who is responsible when an agent buys the wrong item, exceeds intent or is manipulated by a malicious page. Third, watch interoperability. If each network builds a closed agent-payment identity layer, developers will hate it. The market needs standards that make agent authorization portable across wallets, banks and merchants. ## Sources - Visa, 2026: official announcement that Visa and partners completed secure AI transactions. - Mastercard, 2025: Agent Pay announcement and network-level framing for agentic payments. - Stripe documentation: machine payments primitives for automated systems and AI agents. Mentions: Visa, Mastercard, Stripe, Agent Pay, AI agents, agentic commerce, machine payments, tokenization ## Sources - [Visa](https://usa.visa.com/about-visa/newsroom/press-releases.releaseId.21961.html) - [Mastercard](https://www.mastercard.com/news/press/2025/april/mastercard-unveils-agent-pay-pioneering-agentic-payments-technology-to-power-commerce-in-the-age-of-ai/) - [Stripe Docs](https://docs.stripe.com/payments/machine)