Ant International turns payments, FX and treasury into AI-agent infrastructure

Ant International has rolled out an AI-native stack across payments, accounts, foreign exchange, treasury and growth operations, with tools aimed at merchants that are starting to run more of their financial workflow through AI agents. For PSPs and high-risk operators, the interesting part is not the branding; it is that Ant is packaging payment decisioning, FX exposure, liquidity forecasting and account controls into a single agentic layer.

1. Ant International, headquartered in Shanghai and Singapore, said the launch brings together close to 100 products across Alipay+, Antom, WorldFirst and Bettr. The stack is supported by two proprietary foundation models built to handle payment decisioning, FX exposure and liquidity forecasting.
2. The company said it is introducing what it calls the world’s first Account for Agent, a business account built specifically for autonomous AI agents. It is also adding security tooling for agentic commerce, including a dynamic know-your-agent framework linked to its Agentic Mobile Protocol and AgentSafePay, which offers a fund-back guarantee against losses linked to agent-specific risks such as prompt injection.
3. Ant International said adoption is already moving fast inside Antom’s merchant base: 89.5% of Antom merchants have deployed AI agents over the past year, and 81.4% of payment tasks are now completed with AI assistance. For payment providers, that is a sign that agent-triggered payment flows are moving from demo mode into actual merchant operations.
4. The wider network now connects 53 digital wallets and ten national QR code schemes, reaching 150m merchants and more than 2bn consumer accounts across 210 markets, while processing more than 25m transactions each day. Antom 3-in-1 Transformer sits at the centre of the payment stack, with more than ten billion parameters and the ability to process 90 trillion data points annually.
5. Ant International said the model combines behavioural, structured and relationship data to support fraud, scam and abuse prevention while improving payment success rates. Merchants using its newest subscription tier have seen chargeback rates fall by as much as 87%. On the treasury side, FalconTST is designed to forecast FX exposure and cash flow; built on more than 8.5 billion parameters, it is said to achieve accuracy above 93% and reduce hedging and allocation costs by between 30% and 60%.

The Agentic Mobile Protocol went live in April 2026 and already has ten wallet partners and seven acquirers. It also supports settlement between AI agents in increments as small as $0.000001, which tells you where this is heading: not just card-not-present automation, but machine-to-machine payment orchestration at very small ticket sizes.