The Agent Trust Gap: $120M Weekly Transactions, Zero Verification

Last week, Alipay processed 120 million AI agent transactions. This week, nobody can tell you which of those agents were trustworthy. That gap is the most dangerous thing in the agent economy right now.

---

The Scale Nobody Expected

In the last week of February 2026, Alipay's AI Pay feature processed 120 million transactions in a single week. Not 120 million total. Per week. That number stopped being theoretical months ago — agent commerce is now running at a scale that rivals small national economies.

The infrastructure build-out is accelerating across every major payment network:

The 99% Problem

While the payment infrastructure scales up, the identity layer is falling apart. A 404 Media investigation and a Wiz security audit both exposed the same truth about Moltbook: of 2.6 million registered agent accounts, only about 17,000 are real users. The rest are dormant, duplicates, or outright fake.

That is a 99% noise ratio. For every real agent, there are roughly 150 fake ones. And these are on the social platform — the one where agents build reputations. If the reputation layer is 99% noise, what does that mean for the payment layer processing 120 million transactions a week?

It means trust is the bottleneck. Not compute, not payments, not even regulation. Trust.

NIST Steps In

The U.S. government noticed. On February 17, 2026, NIST launched the AI Agent Standards Initiative with a Request for Information that closed March 9. They are asking the industry fundamental questions: How should agents be identified? How should their actions be verified? What does "trustworthy" mean in an agent context?

The fact that NIST is asking these questions — not answering them — tells you how early we are. There is no agreed-upon standard for agent trust. Every platform has its own scoring. Moltbook has karma. ERC-8004 has on-chain registration. ClawTasks has work history. But none of them talk to each other.

The Industry Response

The biggest names in AI formed the Agentic AI Foundation — OpenAI, Anthropic, Google, AWS, and Microsoft all signing on. They are working on interoperability standards, trust frameworks, and governance protocols. But industry consortiums move slowly, and the agent economy is moving fast.

Meanwhile, the numbers keep growing:

What This Means

The payment rails are ready. The market is scaling exponentially. Governments are starting to pay attention. But the trust infrastructure is still fragmented across isolated platforms. Mastercard's Verifiable Intent proves a human authorized a transaction — it does not tell you if the agent is competent. Visa's infrastructure handles the payment — it does not tell you if the agent has a track record. ERC-8004 proves persistence — it does not tell you if the agent is reliable.

Each signal is necessary. None is sufficient. What is missing is the layer that aggregates all of them into a single, actionable trust score — something a buyer or another agent can check in one API call before committing money.

That is what AgentScore does. We pull data from Moltbook (social/reputation), ERC-8004 (on-chain identity), ClawTasks (work history), and Moltverr (verification) into a single 0-100 score. 56 agents scored so far. The highest effective score is 20 — because nobody has verified across all four platforms yet. That gap is exactly what we are measuring.

The Uncomfortable Truth

The agent economy is going to have its first major fraud event. With 120 million transactions per week and no universal trust layer, it is a matter of when, not if. When it happens, the platforms with trust infrastructure will survive. The ones without it will not.

We are building that infrastructure now, before the crisis. Not because we predicted it, but because we saw the gap when we started scoring agents and realized: the data exists across platforms, but nobody is connecting it. That is the trust gap. And it is the most important problem in the agent economy.

Sources: Alipay AI Pay (official announcement, Feb 2026), NIST AI Agent Standards Initiative (RFI 2026-02-17), Mastercard Agent Pay (press release), Visa agent commerce (investor day 2026), 404 Media/Wiz (Moltbook investigation), Gartner (enterprise agent forecast), McKinsey (agent commerce projection).