Fime and Alipay both shipped agentic commerce trust infrastructure on April 21. One built a neutral framework. The other extended a 120-million-transaction platform. The architectural split will decide who owns the layer.

On Tuesday April 21, two announcements landed within an hour of each other. Fime, the French payments certification and standards body, launched FACT, a Framework for Agentic Commerce Trust. Alipay, Ant Group's payment platform, extended its AI Pay service so that any OpenClaw-type agent can initiate payments on a user's behalf. OpenClaw is the shorthand the Chinese market has adopted for terminal-style autonomous agents. Claude Code, Hermes Agent, Alibaba Cloud's JVS Claw, and Ant Group's DTClaw all sit in that category.

Both companies say they solved the same problem. Neither built the same thing.

Fime's pitch is infrastructure. FACT delivers real-time intent validation, policy monitoring, and transaction-level attestation for any payment initiated by an AI agent. Banks get richer authorization data. Merchants gain a way to accept agent-originated transactions. Regulators get real-time oversight they currently lack. The framework is positioned as neutral and interoperable, with no proprietary platform lock-in.

Alipay's pitch is scale. AI Pay already serves more than 100 million users and ran 120 million weekly agent transactions as of February. The April 21 update extends the service to every OpenClaw-type agent. Any of them can install AI Pay and handle purchases in three steps. User states a need. Agent confirms the order. Alipay authorizes payment. No coding is required, and the company says the integration includes multi-layer security on every transaction.

The two launches are not competitors at the product level. A bank cannot buy Alipay. A consumer cannot use FACT directly. They are competitors at the architectural level, and the architectural question is the one that matters.

Fime has spent 30 years running payment certifications. The firm tests EMV implementations, contactless readers, tokenization flows, and digital identity infrastructure. FACT sits in that lineage. The argument is that agentic commerce is the next layer that needs a neutral body to define what safe transaction means, before proprietary platforms lock the market in.

The structural question FACT answers is accountability. When an agent initiates a transaction, who authorized it? What policy governed it? What evidence exists if something goes wrong? Today, those questions resolve through whichever platform the agent runs on. If OpenAI's agent paid, OpenAI's logs are the record. If Alipay's agent paid, Alipay's logs are the record. Neither set of logs is standardized. Neither can be audited across platforms.

Fime's bet is that regulators will eventually require a common attestation layer. The framework exists so that when that requirement arrives, the industry has a neutral answer ready. It is the same bet that produced EMV and 3-D Secure. Both started as industry-agreed standards ahead of regulatory pressure. Both became universal.

The weakness of the standards play is always adoption speed. A neutral framework with no platform pulling it through the market takes years to reach meaningful coverage. Fime has working relationships with card networks and processors, so distribution is plausible. It is not automatic.

Alipay's architecture is the opposite. The user sets up AI Pay inside their preferred agent. Alipay handles intent, policy, authorization, and settlement end to end. There is no neutral layer. There is a vertical stack that works because one operator controls every piece of it.

This is why Alipay can cite 120 million weekly transactions six months into launch. Vertical integration collapses the coordination problem. A merchant does not need to integrate with an agent. An agent does not need to integrate with a merchant. Both parties integrate with Alipay, and Alipay does the glue.

The Chinese context matters here. Alipay's dominant position in merchant acceptance, combined with Ant Group's agent investments, means a single company can ship the full agentic commerce stack in months while Western markets are still arguing over protocols. We covered the China scale dynamic in the Q1 protocol map. Today's announcement widens the lead.

The weakness of the platform play is reach. AI Pay only works where Alipay works. That covers most of China and a growing slice of cross-border commerce through Alipay Plus. It does not cover North American merchants, most of Europe, or any of the existing card network flows. For every market outside Alipay's footprint, the platform model creates a different kind of fragmentation: not too few standards, but too many vertically integrated platforms.

Fime and Alipay are not the only architectures on the board. Earlier this month, FIS launched what the firm called an industry-first offering enabling banks to lead and scale in agentic commerce. The proposition is effectively the platform model, but branded and controlled by banks rather than a technology platform.

Three models are now live. Neutral framework, commerce platform, bank platform. The same month. None of them is definitively winning.

Outside China, agentic commerce is fragmenting fast. Gap built its agent flow inside Gemini by sending checkout to Google Pay. OpenAI pivoted away from Instant Checkout after concluding that rendering merchant-quality checkout inside a chatbot was brittle. Each merchant picked a different platform, which is exactly the fragmentation a neutral framework is meant to resolve and a vertical platform is meant to exploit.

The card networks, which we would expect to be central to any trust-layer outcome, have spent the year hedging. Visa now supports four agentic payment protocols simultaneously. Mastercard has partnered with multiple agent platforms rather than picking one. That hedging is itself a signal. The networks are treating the agentic commerce layer as unsettled. They are keeping options open because they cannot yet tell which architecture wins.

For issuers, the decision that matters is not which framework to endorse today. It is what role the network plays in the layer that eventually wins.

If Fime's model wins, card networks are participants in a neutral standards body. Attestation data flows through the framework. Networks contribute authorization context and receive policy signals. The economic model resembles 3-D Secure. Infrastructure owned collectively, margin earned on the transactions that flow over it.

If Alipay's model wins outside China, card networks become wholesale liquidity providers to a vertically integrated platform. Margin compression is real. Brand presence to the end user is minimal.

If the FIS model wins, banks reclaim the trust layer and the network is the rails underneath. Margins are preserved. Differentiation stays at the card level. But only banks with the scale to operate this layer benefit. The long tail of issuers depends on FIS-like vendors to keep up.

The honest answer is that all three models will coexist for at least three years. The interesting question is which geography sees which model dominate. China is platform. Europe is standards-leaning. The US is fragmenting across bank-branded and platform models. North America has not yet picked.

Three signals before the end of Q2.

First, whether any major US bank adopts FACT as an attestation standard. A top-10 US bank endorsing a neutral framework would tilt the market toward standards before platform lock-in hardens.

Second, whether Alipay announces a North American or European expansion of AI Pay. A press release about Alipay AI Pay inside Claude Code is easy to dismiss as China-only. A partnership with a Western merchant acquirer is a different signal entirely.

Third, whether the next round of card network announcements explicitly backs one of the three models. Visa supporting a fourth standard is hedging. Visa backing FACT, or backing AI Pay integration, or backing FIS's bank model, is a position. Positions tell us more about where the market is going than hedges do.

Every major commerce architecture shift in the past 40 years has been decided by who owned the trust layer. Cards beat cash because the card network was a third-party trust layer. PayPal grew because it was the trust layer between buyers and sellers on eBay. Apple Pay and Google Pay captured mobile payments because they inserted a trust layer between card and merchant.

Agentic commerce will be no different. Either a neutral framework wins and the trust layer becomes shared infrastructure. Or a platform wins and the trust layer becomes a private moat. Or banks reclaim the trust layer and keep card-era economics intact.

We do not yet know which. What we know, as of April 21, is that all three are live.

Standards, platform, or bank-branded. Which trust layer wins?

Charlie Major is a Product Development Manager at Mastercard. The views and opinions expressed in Major Matters are his own and do not represent those of Mastercard.