Is “Know Your Agent” (KYA) Really Necessary? The tale of an Orphan Signal

Short Blog | June 2026

A new category of startup has emerged around “Know Your Agent” (KYA) — the idea that merchants and payment platforms need a framework to verify the identity, authority, and auditability of AI agents acting on behalf of consumers. PYMNTS has covered the space extensively, and KnowYourAgent.xyz is already pitching merchants on “identity, policy controls, and evidence for every AI-agent transaction at checkout.” The framing is intuitive: if a bot is buying something, shouldn’t you know who sent it?

I want to push back — not on the problem, but on whether KYA, as a standalone service category, is the right solution.

Trust Is Domain-Specific — and Risk Is the Point

As I’ve written before in [Open Banking, Open Payments and Trust Networks], the core asset of any trust network is not a list of known participants, it’s a well-defined, enforceable operating model with shared economics. Visa and Mastercard are not trusted because they “know” every actor in the ecosystem. They are trusted because they define roles, certify compliance, enforce standards, allocate liability, and enable Issuers to take on risk when something goes wrong.

This is the question KYA proponents haven’t fully answered: **who takes on the risk, and under what rules?

KYC (Know Your Customer) works because a licensed, regulated institution performs the check, accepts liability for the outcome, and operates under a legal framework that specifies what happens when they get it wrong. A KYA “signal” provided to a merchant is useful only if someone is standing behind it. Without that backstop (governance, liability, enforceability ) KYA is just another signal in a fraud score.

If I’ve Authenticated the Human, and Authorized the Action, What Exactly Am I “Knowing”?

Here’s the question I keep coming back to: if I have perfectly authenticated the consumer and confirmed they authorized the transaction, why do I also need to “know the agent”?

In the current card network model, what flows between banks is a message (ISO 8583 with  operating rules that define what must be done with that message). The card networks don’t “know” every POS terminal or every browser session, Issuers KYC, Acquires Know the merchant and terminals. They certify that the participants operating those endpoints meet their standards, operate within their rules, and accept defined liability for breach.

Google’s AP2 represents the best technical architecture I’ve seen for the agentic era. As I wrote in [Google Rolls Out AP2 and AP2 Operations: Near Term to Long Term the framework uses Verifiable Credentials to authenticate both the actor and the action — enabling the principal hierarchy (consumer → agent → sub-agent) to be cryptographically verified at each step. The architecture is solid. The friction isn’t technical, but rather credential governance: whose credentials are used to sign, where are they stored, and who owns the risk when they’re compromised or misused? That’s a network governance question, not a KYA startup question.

Card Networks Are Already Positioned to Do This

Visa and Mastercard don’t need a third-party KYA layer. They are the trust certification infrastructure. As I’ve argued in [Winning in Network of Networks , networks hold their competitive moat through governance, not technology. A card network can extend their compliance and certification frameworks to say: This agentic platform has been audited, meets our standards, and operates under our rules. Transactions originating from certified platforms carry the same liability framework as card-present transactions.

That’s not KYA. That’s network certification. And it’s far more powerful because it comes with an enforceable operating model and liability backstop — the two things a standalone KYA signal cannot provide.

Where KYA Might Actually Matter

I don’t want to dismiss KYA entirely. There are two narrower use cases where a risk signal about an agent could add value:

1. Merchant-side bot risk scoring. A merchant who receives an order from an unknown AI agent has a legitimate interest in a fraud signal — is this agent well-behaved, or is it a scraper? This is less “Know Your Agent” and more “Agent Reputation Signal,” analogous to IP reputation in fraud scoring. Useful, but not transformative.

2. Unregulated or stablecoin rails. Outside card network governance — stablecoin payments, push payments without chargeback rights — there’s no existing certification framework. Here, a KYA provider could theoretically fill the governance vacuum. But this is a niche that shrinks as card networks extend their agentic certification frameworks.

The Real Problem: Stripe Is Already Solving the Merchant Side

As I covered in Stripe’s Agentic Commerce Protocol (ACP) Stripe isn’t waiting for a KYA ecosystem to emerge. ACP leverages Stripe’s existing device graph and Radar fraud scoring to backfill the device and behavioral signals that traditional checkout generates. Merchants using Stripe don’t need a separate agent identity layer — they’re getting rich device fingerprinting, behavioral signals, and risk scoring baked into the checkout flow. ACP effectively transfers the agent risk problem into the existing merchant fraud infrastructure.

For merchants on Stripe, the question “do I know this agent?” is answered by: “Stripe’s risk models have scored this transaction using the same signals they use for every other checkout.” That’s not KYA — that’s fraud infrastructure doing what fraud infrastructure does.

So What Role Does KYA Actually Play?

Here’s my read: KYA is a genuinely useful concept for the developer and enterprise security community — ensuring that agents running within enterprise systems have defined scopes, audit trails, and permission chains. That’s a real problem worth solving.

But as a payments trust layer? KYA on its own is an orphaned signal without a home. It needs either:

• A network (card scheme, FIDO, AP2 under a governing body) to certify it and enforce it with liability rules, or
• A regulated intermediary to stand behind the assertion

Until one of those two things exists, “Know Your Agent” tells a merchant something interesting — but not something actionable in the way that a network-certified credential or a Stripe risk score is actionable.

The card networks are in the best position to build this. The open question is whether they move fast enough, or whether Google’s AP2 (now donated to the FIDO Alliance) becomes the de facto standard before Visa’s Intelligent Commerce framework fills the same role.

Either way, the startups racing to build KYA as a standalone product are solving the right problem with the wrong business model. Trust requires governance. Governance requires liability. And liability requires a network willing to stand behind it.