September 10, 2026, (Inside AI) — Ant International, Mastercard, and Visa have launched a joint initiative to build a trust framework for AI agents that execute purchases on behalf of consumers. The move targets a fast-emerging gap in digital payments: how to verify that a software agent is authorized to act for a human buyer.
The three payment firms will develop common standards for identifying and authenticating AI agents. The framework would let merchants, banks, and payment networks confirm an agent's identity, permissions, and transaction limits before approving a purchase. No timeline for implementation was disclosed.
The initiative answers a pressing question as autonomous shopping tools spread. Without shared verification standards, each platform could create its own agent credential system, fragmenting the market and raising fraud risk. A common framework could reduce integration costs and speed adoption across e-commerce channels.
Why Agent Identity Is Now a Payment Problem
AI agents that browse, compare, and buy products are moving from demos to production. Payment networks now face a structural challenge: their existing fraud and authentication systems were built for humans using cards or devices. An agent acting on a user's behalf introduces new questions about consent, liability, and transaction limits.
The three firms bring complementary scale. Visa and Mastercard operate the world's largest card networks, processing billions of transactions. Ant International, the global arm of China's Ant Group, runs Alipay+ and other cross-border payment rails. Their combined reach could make the trust framework a de facto standard.
Industry analysts note that agent identity verification is not just a technical issue. It raises policy questions about consumer protection when an agent makes an unauthorized purchase or exceeds a budget. A trust framework could define how liability is assigned and how disputes are resolved.
Sources familiar with the initiative said the first phase will focus on technical specifications for agent credentials. Later phases may cover interoperability testing and certification programs for agent developers. Inside AI could not independently verify these details.
The Race to Define Agent Commerce
This is not the first attempt to standardize agent identity. Earlier efforts in decentralized identity, such as W3C verifiable credentials, have struggled to gain traction in consumer payments. The difference now is commercial urgency: major retailers and marketplaces are already piloting agent checkout flows.
Competing frameworks could still emerge. Cloud providers and AI model developers have their own agent authentication proposals. But payment networks hold a unique position because they sit at the point of transaction settlement. Their standards can be enforced through network rules, not just technical documentation.
The initiative also reflects a broader shift in how payment firms view AI. Rather than treating agents as a threat to card-based revenue, the three companies are positioning themselves as the trust layer for agent commerce. That could preserve their role even if consumers interact less directly with cards or payment apps.
No financial terms or governance structure were announced. The companies said they will engage with regulators, merchants, and technology partners as the framework develops. A formal specification draft is expected within the next year, according to sources.
