Not on every front, and it may not be a race with one winner. Card networks are moving fast on identity, tokens and consumer protection for agents that shop. Crypto rails already carry the tiny machine-to-machine payments cards cannot price: about $73 million across 176 million agent transactions between May 2025 and April 2026, well under a dollar each.
A note on the legal and regulatory points here: they summarise public sources as of November 2026 and should not be relied on. Rules for AI agents are still being written, and some of what is described here is proposed or coming soon rather than settled. Check the current position, or ask a qualified professional, before acting on it.
What have the card networks actually built for AI agents?
Card networks have built ways for a registered, verified agent to pay with a token that stands in for your card, under limits you set. The two largest card networks both launched agent payment programs, the first announced in April 2025.
The designs are similar. Your card number is replaced with a tokenised credential issued for the agent. The agent has to be registered and verified before it can pay. You decide what it may buy. Merchants and banks can see that a transaction was made by an agent, which matters when something is disputed.
In September 2026 the same two networks and a major Asian payments group announced work on a “Know Your Agent” framework. The idea is the card world’s version of know-your-customer checks, applied to software. Its three building blocks:
- Operator traceability: every agent linked to a validated operator, cardholder or business.
- Shared certification: agents assessed against common security and behaviour requirements.
- Continuous monitoring: agents watched through identity and transaction signals.
The aim is to give merchants recognised signals that an agent can be trusted.
Two caveats. Each network keeps its own verification and decision process, so this is coordination, not a single standard. And the announcement said the partners would explore common principles, which means it is early.
Where does crypto have the edge?
Crypto’s edge is price and access at very small sizes: a stablecoin payment can settle a few cents profitably, any time, without a merchant account. That is exactly the shape of most agent spending today.
A crypto investment firm’s report, published in May 2026, counted about $73 million settled by AI agents across roughly 176 million transactions between May 2025 and April 2026. That is well under a dollar per payment. The same analysis found 76% of those transactions fell below a typical $0.30 fixed card fee. On a card, the fee would swallow almost the entire payment.
The plumbing is also open. The x402 standard builds on the web’s “402 Payment Required” status code. A server replies to a request with a price, a network and an address. The agent signs a payment, retries the request, and gets the resource. No card number and no checkout page.
Timing is the other edge. Settlement between banks has traditionally waited for business days. Blockchains do not close. Tellingly, when one of the card networks began settling some obligations in a stablecoin, it cited seven-day availability, including weekends and holidays, as a reason.
How do the two rails compare?
The two rails are built for different jobs: cards for a person delegating a purchase, crypto for software paying software in small amounts. A side-by-side view:
| Card networks | Crypto rails (stablecoins, x402) | |
|---|---|---|
| Built around | A person who delegates shopping to an agent | Software paying software per request |
| Smallest sensible payment | Constrained by fixed per-transaction fees (around $0.30) | Fractions of a dollar; most 2025–26 agent payments were under $0.30 |
| Agent identity | Registration, verification and the emerging Know Your Agent framework | Open registries, still thin (see below) |
| When something goes wrong | Existing dispute processes; agent transactions flagged | No built-in dispute process in the payment standard itself |
| Who can join | Agents approved through the network’s programs | Anyone with a wallet |
| Legal footing | Long-established card payment rules | US federal rules for payment stablecoins since the GENIUS Act was signed into law in July 2025 |
The table shows a trade-off rather than a winner. Cards offer more protection and more gatekeeping. One network said in 2025 it was building a process to help people query agent transactions they do not recognise. Crypto offers lower cost and open access, with less recourse.
Are card networks and crypto competing or converging?
They are increasingly converging: the card world is already using stablecoins, and crypto is borrowing the card world’s identity ideas. The headline race is real, but the two sides are blending.
In December 2025 one of the two networks began letting some US banks settle their obligations with it in a dollar stablecoin. It reported more than $3.5 billion in annualised stablecoin settlement volume as of 30 November 2025. The person tapping the card sees no difference; the money moves underneath on a blockchain.
Meanwhile crypto’s answer to “which agent is this?” is still weak. A June 2026 academic study of ERC-8004, an open standard for on-chain agent identity and reputation, found that only 3% of registrations on Ethereum, and 4% and 15% on the two other networks it studied, had a valid registration file with a live service. It flagged 73.5%, 59.2% and 90.6% of reviewers on those three networks as showing coordinated fake-account behaviour. The card networks’ Know Your Agent work targets the same problem from the other side.
To be fair, the open approach has real strengths. ERC-8004 sets out three public registries: one for identity, one for reputation and one for validation of an agent’s work. Anyone can inspect them, and no network has to approve an agent. The study’s finding is that, as deployed so far, the reputation layer cannot yet be relied on as a trust signal.
Our judgement: the likeliest near-term picture is split. Cards handle agents buying things for people at checkout. Stablecoins handle agents paying other services per request. Whoever solves identity and accountability first gains ground on the other’s turf.
What could go wrong?
On the card side, the risk is gatekeeping. If an agent must be certified by a network to pay, the networks decide which agents exist in commerce. That brings safety, and it also concentrates control.
On the crypto side, the risks are thin identity and concentration. Fake reputation is cheap on open registries, as the ERC-8004 study showed. And the same May 2026 analysis found more than 98% of agent settlements ran through a single dollar stablecoin. In our view, that makes one issuer a single point of failure for most of today’s agent payments.
On both sides, the market is small and people are wary. $73 million in a year is modest next to global card volumes. And a major research firm reported in June 2026 that three-quarters of online adults in the US, UK and Canada were uncomfortable letting an agent pay on its own. Neither rail has yet won the person whose money it is.
Finally, beware of big forecasts. The September 2026 announcement cited a projection that agents could orchestrate $3 trillion to $5 trillion of consumer commerce by 2030. That is a projection by a consulting firm, not a measurement.
The trust takeaway
The card-versus-crypto question matters less to you than a simpler one: when your agent pays, what protects you? Card rails lean on identity checks and dispute processes. Crypto rails lean on limits you set in advance, because recovery afterwards is harder.
Before you let an agent pay on either rail, ask:
- Which rail is it using? A card token, a stablecoin, or both?
- Who verified the agent? Is it registered with a network, or only listed in an open registry?
- What happens if it pays the wrong party? Is there a dispute process, or is the payment final?
- What limits apply? Per-payment caps, totals, approved merchants or services.
- Who is accountable? The agent’s maker, the platform, or you?