---

Discussion

Sam Dewinski:

The new element is interoperability. Payment networks have long used private technical rules to coordinate banks, merchants and processors. But an AI agent adds another intermediary: software that can interpret an instruction, choose among options and act without a person approving every checkout screen. Visa, Mastercard and Ant International already had separate approaches. Visa has its Trusted Agent Protocol, Mastercard has agent tokens and Verifiable Intent, and Ant has a mobile-focused protocol. This collaboration is an attempt to connect those systems rather than start with one entirely new standard.

Kate Burvish:

And that connection is where the economic promise comes from. A merchant does not want a separate trust relationship with every shopping agent in existence. A common signal could reduce duplicated integrations and screening costs, while giving participants more visibility into transaction risk. But those are the companies’ expected benefits, not measured results. There is no final specification, adoption timetable or evidence yet that fraud or integration costs have fallen.

Ann Tofado:

Which is why the project’s status matters. BuildFin.ai has a public-policy connection because Singapore’s Monetary Authority convened it, but the announcement does not say that the authority has endorsed this framework or will require anyone to use it. For now, this is private-sector standard-setting with regulatory relevance. If it becomes widely accepted, it could shape practical rules about identity and authorization before governments settle questions of liability and consumer protection.

Red Velhouse:

So the headline is not that a new payment rule is already in force. It is that competing systems are trying to recognize one another. Let’s make that concrete. “Know-Your-Agent,” or KYA, has to answer more than whether a piece of software is genuine. Who operates it? Which user or business does it represent? What did that user authorize? And did the agent stay within those limits? Sam, where does identity end and authorization begin?

Sam Dewinski:

Identity is only the first layer. Visa’s protocol uses cryptographic signatures to help a merchant verify an agent, its session and its claimed authorization. Mastercard’s Verifiable Intent is aimed at creating a tamper-resistant record of what the user authorized. Those are related but different tasks. Proving that an agent is authentic does not automatically prove that a particular purchase followed the user’s instructions. Historically, payment authentication often asked whether the payment instrument was valid. Agentic commerce also asks whether the decision itself was valid.

Kate Burvish:

And that difference creates the potential for costly disputes. An authenticated agent might buy the wrong product because it misunderstood a broad instruction, or because the product information was misleading. The transaction could be cryptographically genuine and still be economically wrong for the consumer. The announcement does not say how losses would be divided among the user, agent platform, merchant, wallet or card network. Until that is clearer, businesses may hesitate to let agents make consequential purchases.

Ann Tofado:

That is also where a technical record stops being a complete political solution. A record of authorization may help establish what happened, but it does not decide who should bear the loss when an agent exceeds its mandate, follows manipulated instructions or buys from a deceptive seller. Private network rules can govern their participants; they do not necessarily settle consumer rights across borders.

Red Velhouse:

Kate, if interoperability is supposed to lower barriers, could it also give the biggest payment and technology firms more control over commerce?

Kate Burvish:

Yes. Common trust signals could help smaller merchants avoid integrating with dozens of agents. But if participation depends on certification programs, monitoring systems or proprietary credentials controlled by a few large firms, compliance becomes a new fixed cost. Large platforms are better positioned to absorb that cost. Smaller agents could face disproportionate burdens even if the framework is described as open. Technical interoperability does not necessarily mean equal bargaining power.

Sam Dewinski:

That is the historical tension. Standards can connect participants while also reinforcing the position of the networks that administer them. Visa and Mastercard are trying to make their identity, token and authorization systems relevant to a new commercial layer. At the same time, Ant’s Agentic Mobile Protocol is aimed at wallets, super apps and smart devices, not only card checkout. So this is both an interoperability effort and a competition among payment ecosystems.

Ann Tofado:

And those two things can happen at once. Participants may genuinely need shared rules because fragmented protocols would make agentic commerce cumbersome. But whoever controls certification and liability can influence which agents, merchants and wallets are considered trustworthy. If the framework becomes a de facto rulebook, it could affect competition without ever becoming a law. That is why access conditions and independent oversight matter.

Red Velhouse:

Let’s stay with that question of who gets to write the rulebook. Ann, should governments stop private networks from moving first, or is there a useful role for experimentation?

Ann Tofado:

Private experimentation can move faster than formal regulation, and Singapore may be using BuildFin.ai as a venue for responsible financial-AI development. But governments should not confuse convening with approval. They can let technical work proceed while setting baseline expectations for consent, data use, security testing, revocation and dispute resolution. Otherwise, a voluntary framework could become practically mandatory because merchants and smaller providers cannot afford to operate outside it.

Kate Burvish:

There is a related data issue. KYA could improve accountability by linking an agent to an operator, cardholder or business and monitoring identity and transaction signals. But those signals could also become valuable repositories of behavioral and purchasing information. The commercial incentive will be to use them for fraud prevention, ranking, personalization and possibly competitive advantage. Data minimization and limits on cross-network access may determine whether the system reduces risk or concentrates informational power.

Sam Dewinski:

And the history of tokenization offers a useful warning. Tokens and signatures can make transactions traceable and harder to forge, but they cannot tell us whether a recommendation was biased, a seller’s description was deceptive or a person understood the consequences of a broad instruction. An agent is not merely a payment credential. It interprets and makes choices. That is the important difference from older forms of automation.

Red Velhouse:

That brings us to the market itself. Ant International has cited projections that agents could orchestrate three to five trillion dollars in consumer commerce by 2030. Is that evidence of an emerging market, or mainly an incentive to build the infrastructure now?

Kate Burvish:

It is a signal of expected opportunity, not proof that the market will reach that size. Agents could reduce search and checkout friction, and payment companies want to preserve their role as commerce shifts toward software interfaces. But consumers may delegate routine purchases long before they trust agents with expensive or consequential ones. The evidence to watch is repeat consumer use, merchant adoption, dispute rates and measurable savings—not just demonstrations or protocol launches.

Ann Tofado:

There is a legitimacy question inside those economic choices. If an agent selects products, ranks sellers and authorizes payment, a platform may influence several stages of commerce while presenting the outcome as the user’s choice. Regulation will have to consider not only whether someone authorized the purchase, but how the agent reached its recommendation and whether the user can inspect, revoke and contest its actions. Those questions become harder when a purchase crosses countries, wallets and marketplaces.

Red Velhouse:

We have identified the main unresolved pieces: identity, valid intent, liability, data and access. What would show that this collaboration is becoming real infrastructure rather than a well-publicized alignment among incumbents?

Sam Dewinski:

First, a published technical specification explaining how the existing protocols connect. Second, evidence that independent agents and different payment ecosystems can participate, rather than only the three announcing companies. Third, security testing that examines compromised agents, manipulated content and the gap between authentic identity and valid intent. A standard gains historical importance when others can rely on it, not merely when its creators describe the ambition.

Kate Burvish:

I would add a transparent cost and liability model. Who pays for certification? Who bears losses after a compromise or misunderstanding? Can merchants avoid multiple overlapping programs? If the framework lowers integration costs for smaller participants, that should appear in adoption patterns. If it creates another tollbooth, the economic result will be consolidation presented as interoperability.

Ann Tofado:

And regulators should watch whether private rules become unavoidable without public accountability. That does not require governments to design every technical detail. It does require baseline rights: meaningful consent, access to records, the ability to revoke an agent, a route to challenge unauthorized purchases and safeguards around identity and behavioral data. The political question is whether those rights are built in before the infrastructure becomes too entrenched to change.

Red Velhouse:

So the immediate story is modest but consequential. Visa, Mastercard and Ant International have begun work on a shared Know-Your-Agent approach, building on competing systems for agent identity, tokens, intent and mobile payments. The unresolved issue is whether interoperability will distribute trust more broadly—or put a new layer of commerce under the control of a small number of networks and platforms. The next signals are a real technical specification, independent security evaluation, merchant and wallet adoption, regulator guidance, and a clear chargeback or dispute regime. Those developments will show whether agentic commerce is becoming dependable infrastructure or remaining a heavily promoted experiment.

Red Velhouse:

The central unresolved issue is accountability: when an AI agent acts with valid credentials but makes an invalid or harmful decision, who answers for it? The next signals are a published specification, independent security testing, real merchant adoption, regulator guidance and clear rules for disputes and losses. Sources and references for this discussion are available with the episode at Factolio.com.

---

Sources and References

These sources supported the factual material used in this discussion. Factolio’s panel discussion is AI-generated from researched evidence and is written in original language.

1. Reuters — Payment firms Visa, Mastercard and Ant International team up on AI agent trust framework (NEWS)
2. Ant International — Ant International, Mastercard and Visa Initiate Collaboration on Know-Your-Agent Interoperability to Scale Agentic Commerce (PRIMARY)
3. Visa — Visa Unveils Trusted Agent Protocol for AI Commerce (PRIMARY)
4. Visa Developer Center — Trusted Agent Protocol (PRIMARY)
5. Mastercard — Agentic token framework: Driving trusted AI transactions (PRIMARY)
6. Mastercard — How Verifiable Intent builds trust in agentic AI commerce (PRIMARY)
7. Ant International — Ant International Launches Open-Sourced Agentic Mobile Protocol to Drive AI Commerce (PRIMARY)
8. Visa — Visa Core Rules and Visa Product and Service Rules, April 2026 edition (PRIMARY)
9. Associated Press — Visa brings payments to ChatGPT as AI agents start buying for you (NEWS)
10. arXiv — Beyond the Mandate: A Systematic Security Analysis of the Agent Payments Protocol (ANALYSIS)