Visa, Mastercard and Ant International Just Agreed on How to Trust AI Agents: What Know Your Agent Means for Payment Providers
TL;DR
Visa, Mastercard and Ant International are developing an interoperable Know Your Agent (KYA) framework for AI agents that initiate payments for consumers and businesses.
The objective is straightforward: an AI agent verified through one participating provider should be recognisable across other payment ecosystems without repeating the entire registration process.
The framework will focus on:
Identifying the AI agent and its operator
Linking the agent to a consumer or business mandate
Sharing certification and trust signals
Continuously monitoring agent behaviour and transactions
Preserving each network’s own risk, authorisation and settlement decisions
This is an important shift for payment providers. The next competitive battleground is not only the payment rail. It is the agent identity rail that determines which AI agents can access merchants, wallets and payment credentials.
The collaboration, announced on 10 September 2026, will run through BuildFin.ai, a platform convened by the Monetary Authority of Singapore and built on its SAFR framework.
What does Know Your Agent mean for payment providers?
Know Your Agent means payment providers will need to verify not only who is paying, but which software agent is acting, who operates it and what authority it has.
Traditional payment controls were designed around a human cardholder, merchant and transaction. Agentic commerce introduces another participant: software that can discover products, make decisions, initiate transactions and potentially repeat them.
KYA extends the logic of Know Your Customer into this new environment. It is designed to answer four questions:
What is the agent?
Who operates or controls it?
Which consumer or business has authorised it?
Is the transaction within the agent’s permitted scope?
The Visa, Mastercard and Ant International initiative will establish common principles for this process while allowing each network to retain its own verification and decision-making systems.
The announcement centres on three areas:
Operator traceability: every agent should be linked to a validated developer, platform, business, cardholder or account holder.
Shared certification: agents should meet baseline security and behavioural requirements before accessing payment ecosystems.
Continuous monitoring: certification should not be a one-off event. Agent behaviour, transaction patterns and risk signals will need ongoing assessment.
That last point is particularly important. An agent can be legitimate when it is registered and still become compromised, manipulated or misconfigured later.
The big change is from payment rails to agent identity rails
Payment companies have spent years building faster, cheaper and more flexible ways to move money. AI agents now create a different infrastructure challenge: deciding whether an automated actor should be allowed to begin the payment in the first place.
Visa has developed the Trusted Agent Protocol. Mastercard has developed Verifiable Intent, which is designed to prove that an AI action reflects a user’s authorised instructions. Ant International has developed its Agentic Mobile Protocol.
These approaches are not being replaced by one universal payment rail. Instead, the companies are exploring ways for their identity, authorisation and trust signals to be recognised across networks.
That distinction matters.
A payment rail answers: How does the transaction move?
An identity rail answers: Should this agent be allowed to initiate the transaction, and under what conditions?
For fintechs, payment facilitators, processors and gateways, this creates a new strategic layer between the merchant interface and the authorisation request. The ability to recognise a trusted agent may become as important as tokenisation, fraud scoring and routing.
Why interoperability is the competitive battleground
Agentic commerce will not develop inside a single network or wallet ecosystem. Consumers may use one AI assistant, merchants may accept several payment methods, and the underlying transaction may involve a card issuer, wallet, acquirer, payfac and multiple risk providers.
Without interoperability, every agent platform would need to register separately with every network and payment provider. That would create duplicated onboarding, higher compliance costs and slower adoption.
Jiang-Ming Yang of Ant International described the principle directly: if an agent registers with Ant, it should not need to register again with Visa or Mastercard. “Trust is the foundation of the AI transformation.”
Mastercard Chief Digital Officer Pablo Fourez made the commercial implication equally clear: “Interoperability across Know-Your-Agent frameworks is essential to making agentic commerce work at scale.”
The competitive question is therefore changing from:
Which network owns the agent payment?
to:
Which networks and platforms provide the most widely accepted trust signals?
The provider that becomes easiest for merchants, wallets and agent platforms to connect to may gain distribution even if it does not own every transaction.
This is similar to the role payment orchestration plays today. Businesses want one integration that can access multiple rails. In agentic commerce, they will increasingly want one integration that can recognise multiple agent credentials, consent models and protocol formats.
What this means for merchant acceptance strategies
Merchants will need to make themselves readable and acceptable to AI agents, not just easy for humans to navigate.
Mastercard’s Agent Connect, launched on 9 September 2026, is designed to provide a connection point for AI shopping agents to discover catalogues, build carts and initiate payment. Visa’s Intelligent Commerce Connect is pursuing a similar acceptance-layer role across agentic protocols and payment environments.
For merchants and POS companies, this points to several practical changes:
Product catalogues will need structured, machine-readable data.
Pricing, availability, delivery and tax information must be available in real time.
Checkout flows must accept verified agent credentials.
Offers and loyalty rules will need to be understandable by software.
Merchant systems must distinguish trusted agents from malicious automation.
Receipts and transaction records should retain relevant intent and delegation evidence.
POS providers may also need to support agent-originated transactions in physical environments. A consumer could ask an agent to purchase a product for collection, arrange a recurring service or complete a transaction through a wallet at the point of sale.
The merchant that accepts trusted agents efficiently may gain visibility in a new discovery channel. The merchant that blocks all automated traffic may become invisible to the next generation of shoppers.
Fraud and risk operations will need to monitor the agent, not just the payment
KYA will not eliminate fraud. It will give fraud teams more context to work with.
Risk operations will need to monitor:
Agent identity and certification status
Operator history and linked merchants
User mandates and spending limits
Device, credential and token behaviour
Changes in agent software or permissions
Transaction velocity and unusual purchasing patterns
Conflicts between declared intent and actual behaviour
Repeated declines, refunds, disputes or account takeovers
This is a move from static verification to continuous agent assurance.
A trusted identity should not create automatic approval. Each payment provider will still need to make its own decision based on transaction risk, merchant category, geography, customer profile and available funds.
For issuers, Mastercard’s Agent Pay model and Verifiable Intent point towards a future where the issuer can assess not just the amount and merchant, but the authority behind the instruction. Was the agent told to buy a particular product? Was there a spending limit? Was the mandate still valid? Did the final order materially change?
These signals could improve fraud detection and dispute handling, but they also create new obligations around logging, privacy, explainability and consumer support.
Wallets are central to the agentic commerce opportunity
Digital wallets are already a major part of global commerce. According to Worldpay’s Global Payments Report 2026, wallets represented 56% of global e-commerce transaction value and 33% of global point-of-sale transaction value in 2025.
That gives wallet providers a strong starting position. Wallets already combine identity, authentication, payment credentials, loyalty, merchant relationships and user permissions in one environment.
Ant International’s Alipay+ connects 150 million merchants to more than 50 digital wallets and banking apps. Its scale means KYA interoperability is not only a technical standards exercise. It is also a contest to make wallet ecosystems usable by AI agents across borders.
Alipay’s current examples show how this may develop. Users can automate recurring Starbucks orders and Didi ride-hailing through an AI assistant, while retaining payment approval. The AI handles the scheduling and service interaction, but the user still authorises the payment.
That model is commercially useful because it separates automation of the task from delegation of the money movement. Over time, some users may permit transactions within predefined limits. Others may require approval for every payment.
Wallet providers should prepare for both behaviours.
What each payments business should do now
Fintechs and start-ups
Build agent identity, consent and auditability into the product architecture from the beginning. Do not treat KYA as a compliance feature to add after launch.
Prioritise:
Agent registration and credential management
Operator verification
User mandate and permission controls
Protocol translation
Detailed transaction logs
Clear revocation processes
POS companies and merchants
Make catalogues, inventory, prices and checkout processes machine-readable. Assess how your platform will recognise a certified agent and how it will record the user’s authority.
Payfacs, ISOs and payment facilitators
Update underwriting and monitoring models to account for agent operators, agent platforms and delegated transaction flows. Your merchant risk model may need to evaluate the agent ecosystem connected to a merchant, not only the merchant itself.
Payment processors and gateways
Develop an agent acceptance layer that can support multiple protocols. The commercial opportunity may be less about owning one protocol and more about making complex agent payments simple for merchants.
Remittance businesses
Expect agents to compare providers, fees, exchange rates and delivery options on behalf of customers. Agent identity and beneficiary protection will be essential, particularly where transactions cross borders and involve high-risk corridors.
Card issuers
Prepare for new authorisation data, including agent identity, user intent, mandate scope, validity period and operator reputation. Issuers should also define how customers can pause, revoke or dispute agent permissions.
A practical KYA readiness checklist
Payments businesses should ask:
Can we identify every AI agent that reaches our payment environment?
Can we trace that agent to an operator and an authorised user or business?
Can we verify whether the agent is certified under a recognised framework?
Can we interpret credentials from more than one agent protocol?
Can we enforce merchant, amount, category and time-based permissions?
Can we continuously monitor agent behaviour after onboarding?
Can we revoke an agent without replacing the customer’s entire payment credential?
Can our dispute team access a reliable record of user intent?
Can our merchants accept trusted agents without building separate integrations?
Can our privacy and compliance controls handle shared trust signals across networks?
The strategic conclusion
The Visa, Mastercard and Ant International announcement is not simply about adding AI to payments. It is about creating a common trust layer for software that will increasingly act inside payments.
McKinsey estimates that AI agents could orchestrate US$3–5 trillion of global consumer commerce by 2030. That opportunity will not scale if every agent must repeatedly prove itself to every network.
Visa’s Rubail Birwadker said, “As AI agents become a bigger part of how people discover and buy, trust must scale with them.”
That is the central message for payment providers. Agentic commerce will need payment rails, but it will first need identity rails. The companies that help merchants, wallets, issuers and platforms recognise trusted agents across ecosystems will be best placed to shape the next phase of payments.
For businesses planning that transition, RivaTech Consulting’s payment strategy and orchestration expertise can help connect the commercial, operational and technical decisions required for agentic commerce.
Frequently asked questions
What is Know Your Agent in payments?
Know Your Agent is a framework for identifying, verifying and monitoring AI agents that initiate transactions on behalf of consumers or businesses. It links the agent to an accountable operator and an authorised user or organisation.
Why are Visa, Mastercard and Ant International developing an interoperable KYA framework?
Interoperability can reduce duplicated agent registration and allow trusted signals to be recognised across card and wallet ecosystems. This should lower integration costs and support faster adoption of agentic commerce.
Does KYA replace payment network risk controls?
No. The framework is intended to provide common identity and trust signals while Visa, Mastercard, Ant International and other providers retain their own verification, risk, authorisation and settlement decisions.
What should payment providers do to prepare for AI agents?
They should build agent registration, operator traceability, user consent, delegated payment controls, continuous monitoring, protocol interoperability and auditable intent records into their payment architecture.