Google backs AP2 payment protocol

Payments are the hard part of AI shopping. Getting a chatbot to find a flight or compare shoes is easy enough. Letting that chatbot actually spend your money is where things break — because card systems assume a human clicked “buy,” and an autonomous agent blows up that assumption. That is the gap Google and Mastercard are trying to close. This week, Google said it is donating AP2 — its Agent Payments Protocol — to the FIDO Alliance, and Mastercard said it is donating Verifiable Intent alongside it. ### What is AP2, exactly? AP2 is an open protocol for agent-led payments. Basically, it gives AI agents a standard way to carry payment instructions, user permissions, and transaction details across platforms, instead of every wallet, merchant, and model inventing its own rules. Google introduced AP2 in September 2025 as a way to handle authentication, authorization, and accountability when an agent acts for a user. ### Why wasn’t that enough on its own? Because “my agent bought this for me” is not the same thing as “I clearly authorized this exact purchase.” If a bot books the wrong hotel, exceeds a spending cap, or gets tricked by a malicious merchant, everyone involved needs a shared record of what the user allowed. AP2 handles the payment flow. But the ecosystem also needs proof of intent — not just proof that a transaction happened. ### So what is Verifiable Intent? It is a cryptographic trust layer for agentic commerce. Mastercard and Google introduced it in March 2026 as an open-source framework that creates a tamper-resistant record of what the user approved when an AI agent acts on their behalf. Think of it like a signed instruction sheet attached to the transaction — budget, merchant constraints, timing, and other permissions — so merchants, issuers, and wallets can all check the same facts later. ### What changed this week? The big move is governance. Google is not just supporting AP2 in its own products anymore — it is handing the protocol to FIDO, the industry group best known for passkeys, so the standard can be developed in a broader, interoperable forum. Google also said Verifiable Intent, co-developed with Mastercard, is being donated to FIDO too. That turns two related building blocks into a shared standards project instead of a pair of company-led specs. ### Why FIDO? Because FIDO already lives in the world of digital trust. Passkeys solved a similar problem for logins — how do you prove a real user approved access without relying on weak, replayable signals like passwords? Agentic payments need a version of that for commerce. FIDO’s pitch is that AI agents should carry passkey-grade proof when they act, not vague claims that the user “probably meant it.” ### Who else is behind this? Google said AP2 was originally developed with a broad group of payments and tech partners, including Mastercard, PayPal, Adyen, and Worldpay. That matters because a payments standard dies fast if merchants, wallets, issuers, and processors do not all recognize it. The whole point is interoperability — one agent should not need a different trust model for every checkout stack on the internet. ### What problem are they really trying to stop? Fraud, disputes, and chaos. If AI agents start buying things at scale, the ugly cases will pile up fast — unauthorized purchases, unclear liability, fake agent instructions, and endless chargeback fights over who approved what. A tamper-resistant audit trail will not eliminate bad behavior, but it gives every party a common record to inspect. That is a much better starting point than screenshots and shrugging. ### What’s the bottom line? This is early infrastructure, not mass-market checkout yet. But it is a real signal that AI commerce is graduating from demo territory into standards work. Google is pushing AP2 beyond its own orbit, Mastercard is adding a proof-of-intent layer, and FIDO is becoming the place where those rules may get hammered into something the wider payments industry can trust.