For years the card networks treated AI-agent payments as a laboratory experiment. That phase appears to be over. Visa is now working with Mastercard and Ant International on shared standards for verifying autonomous purchasing — common rules for how a machine acting on a human’s behalf proves who it is, what it is authorized to buy, and how liability is assigned when something goes wrong. The timing matters more than the press release: this is the incumbents moving from observing agent commerce to defining its rulebook.
What the Standards Are Actually For
The core problem in agentic payments is not moving value — stablecoins and card rails can both do that — it is verification. When an AI agent initiates a purchase, the merchant needs to know the agent is legitimate, the human principal actually delegated this authority, and the transaction falls within the scope of that delegation. Card networks have spent decades building exactly this kind of infrastructure for human cardholders: tokenization, identity attestation, dispute resolution, chargeback rules. What Visa, Mastercard and Ant are reportedly standardizing is a machine-readable equivalent — how an autonomous purchaser gets credentialed, and how the networks treat those credentials interchangeably rather than as proprietary silos.
The strategic logic is straightforward. If verification standards for agents are set by the networks, then agent commerce defaults to card rails regardless of what settles underneath. That is a direct answer to the stablecoin-native approach, where protocols like x402 and Coinbase Agent Payments assume machine identity and wallets live on open infrastructure, not inside a network’s rulebook.
The Skeptic’s Case: Overvalued Before It Exists
Not everyone buys the narrative. Simply Wall Street’s analysis flags Visa as potentially as much as 88% overvalued, with AI payments and onchain credit cited as among the expectations already priced into the stock. The substance of that call is debatable, but the underlying observation is not: agent-initiated transaction volume today is trivially small. Networks standardizing verification for a category that barely registers in dollar terms is either early-stage positioning or evidence that the agentic-payments narrative is running well ahead of usage. Both readings are plausible simultaneously.
There is also a real possibility that the standards work is defensive. If autonomous purchasing eventually scales on stablecoin rails, the networks’ toll position erodes. Writing the verification layer now — and making it interoperable across Visa, Mastercard and Ant’s Alipay ecosystem — is a way to keep agents inside a permissioned perimeter.
The Demand-Side Reality Check
S&P Global’s Q2 2026 consumer survey puts hard numbers on the adoption gap. Stablecoin awareness among U.S. consumers sits at just 16%, and 47% of respondents said they would not let an AI make purchases on their behalf. Standards for verifying agents do nothing, by themselves, to change either number. The binding constraint on agentic commerce right now is not verification plumbing — it is whether ordinary consumers will delegate spending authority at all, and whether regulators will let them.
Meanwhile, the governance layer is being built in parallel. At Money20/20 Asia, MetaComp introduced what it calls the first AI agent governance framework for regulated financial services, anchored in its StableX know-your-agent tooling. Verification standards from the networks and KYA frameworks from regulated institutions are converging on the same question: who vouches for the machine?
What to Watch
The open questions are concrete. Will the Visa-Mastercard-Ant standards be published openly or licensed, and will stablecoin-rail protocols adopt them or treat them as competitive exclusion? Does interoperable agent verification require agents to hold network credentials, effectively re-intermediating the networks into machine commerce? And will regulators — several of whom are already circling agent-initiated payments — accept network-run verification as sufficient consumer protection, given that nearly half of consumers say they do not want agents spending their money in the first place?
The most likely near-term outcome is coexistence: card rails for high-value delegated purchases where dispute rights matter, stablecoin rails for sub-cent machine-to-machine settlement where they do not. The standards fight is about everything in between.
Sources
- https://simplywall.st/stocks/us/diversified-financials/nyse-v/visa/news/visa-v-could-be-88-overvalued-on-ai-payments-and-onchain-cre
- https://thefintechtimes.com/sp-global-us-consumers-wary-of-stablecoins-and-agentic-ai/
- https://www.openpr.com/news/4630102/best-crypto-to-buy-now-visa-and-mastercard-push-ai-payments-as
- https://yellow.com/001thm%3Bzzzziwwww.jxpta.comm/press-releases/metacomp-launches-the-world-s-first-ai-agent-governance-framework-for-regulated-financial-services