Ant International has launched an open-source payment protocol backed by major wallets, acquirers, Visa and Mastercard as rising demand for AI shopping agents collides with consumer fears over autonomous spending.
Software agents are moving beyond chat-like assistance and towards a role in commerce itself, but the shift is still constrained by a basic problem: trust. Global Payments research found that consumers now expect AI-powered agents to handle 15% of their purchases within five years, up from 9% a year earlier, while 45% of Americans said they have used or would consider using an AI shopping agent. Yet the same research also found that 50% remain worried about payment security, a reminder that enthusiasm for AI discovery does not automatically translate into comfort with autonomous spending.
That gap helps explain why payment companies are racing to build infrastructure designed specifically for agentic commerce. Ant International has unveiled the Agentic Mobile Protocol, or AMP, as an open-source framework intended to connect AI agents to mobile payment systems such as digital wallets and super apps without stripping users of control over their accounts. According to Ant International, the protocol is built around four core elements: Know-Your-Agent checks, permissioned delegation, wallet connectivity and agent-to-agent settlement.
The company is not developing the system in isolation. Ant International says AMP’s first rollout is supported by 10 Alipay+ wallet partners, representing 1.5 billion user accounts, and eight acquirers: Adyen, Allinpay, Antom, Checkout.com, Fiserv, Global Payments, Nuvei and Worldline. It has also started working with Visa and Mastercard on a Know-Your-Agent interoperability framework, an attempt to establish common identification principles for AI agents while allowing each payment network to retain its own verification processes.
The trust problem is not limited to consumer transactions. Ant International’s chief innovation officer, Jiang-Ming Yang, said at the Voyage conference that he expects adoption to move faster in business-to-business settings than in consumer markets, where users may be slower to hand over decision-making to software. That makes sense: enterprise workflows often follow clear rules, fixed budgets and measurable objectives, which can make it easier to justify the use of agents for tasks such as software subscriptions, advertising spend and procurement.
AMP’s architecture is designed to reduce the friction that still makes conventional card payments awkward for autonomous software. The protocol is meant to turn an agent’s validated request into a payment instruction without exposing card details, bank credentials or private application programming interfaces. Ant International says the system can cut the steps needed to link a payment agent to a digital wallet by half compared with traditional card-binding methods, while leaving the wallet in charge of authentication triggers, spending limits and other controls.
The protocol also tries to answer one of the hardest questions in agentic finance: who carries the risk when an autonomous system gets it wrong? Global Payments research showed that half of consumers cite payment security concerns, while a third want explicit approval for each transaction. Ant International’s answer is to offer two modes of use. In one, the user stays present during checkout and gives the final approval. In the other, the user grants a bounded task in advance, with the ability to monitor, change or revoke it in real time. The company has also introduced AgentSafePay, which it describes as a money-back guarantee for merchants against risks specific to agent-driven transactions.
Beyond checkout, Ant International is positioning agentic tools as part of a wider operating stack for international business. The company says it has launched AI-native tools across payment, account management, foreign exchange, treasury and growth operations. Among them is FalconTST, an AI model for FX and cash-flow forecasting that Ant says has more than 8.5 billion parameters and more than 93% accuracy in real-world forecasting cases, while also reducing hedging and allocation costs by 30% to 60%. Ant International has also said its 3-in-1 transformer model has helped it replace more than 200 legacy point solutions, and that its merchant-facing tools, including Antom Autopilot and SmartDispute, are designed to automate payment operations, dispute handling and market expansion.
The larger commercial case for this technology is clear enough: if AI agents are going to buy, negotiate and settle transactions on behalf of people and businesses, the payment layer has to be rebuilt around them. But the industry still has to solve a more difficult question than software engineering alone. Agents need recognised identities, clear limits on what they are allowed to do and a way to carry trust across different payment rails. AMP is Ant International’s attempt to build that foundation, but its real test will be whether rivals, regulators and merchants agree on the same rules.
Disclaimer: This article is intended to inform and educate, not to recommend or endorse any financial product, investment or strategy. Please consider your own financial circumstances and seek professional advice where appropriate before making financial decisions.
