AI agents are learning to shop. Alipay wants to help them pay, too.
Ant Group has launched two Alipay tools for agent-driven commerce: AI Wallet, a consumer control layer inside the Alipay app, and Token Pay, a settlement product for AI model providers. The tools are built for AI agents that can shop, book services, and complete transactions while giving users and payment providers more control over authorization and risk.
Alipay adds controls for agent shopping
According to the South China Morning Post (SCMP), Ant Group described the rollout as a “full-stack” AI payment solution for AI model developers, retail businesses, and consumers.
AI Wallet is the consumer-facing part. It lets Alipay users monitor, manage, and authorize tasks carried out by AI agents before, during, and after a transaction. Token Pay is built for AI model providers to handle subscription models, in-agent token top-ups, and complex micro-transactions.
SCMP reported that Ant Group CEO Cyril Han Xinyi described the logic behind the system this way: “Agents execute payments, while tokens carry value.”
AI Wallet is about user permission; Token Pay is about payment flows between agent platforms, model providers, and merchants. Together, they show how AI shopping agents could move from recommendations to transactions.
The example Ant Group gave is simple enough to understand. An AI agent running on smart glasses could order, customize, and pay for coffee before the user enters a café. The user avoids checkout friction, but the payment system still needs to know who approved the transaction, what the agent was allowed to do, and when the user can intervene.
That is the same problem behind agentic commerce: once an AI agent can move from product discovery to purchase execution, payment controls become part of the core infrastructure.
Alipay already has a large base for that push. Ant Group said in February that Alipay AI Pay had surpassed 100 million users and processed more than 120 million transactions during the week of February 5-11, 2026. In April, Alipay also launched an AI Pay service for OpenClaw-type agents, with identity verification, user authorization for each payment, 24/7 risk controls, and account-protection coverage.
Payments teams need new checks
Accenture estimated that more than 30% of online commerce could run through AI agents by 2030, representing nearly $3.1 trillion in transactions. But its payments research found that only 12% of consumers are willing to let agents handle payment choices today.
Fraud teams are also under pressure. Accenture found that 78% of financial payments leaders expect fraud to increase significantly with agentic payments, while 87% see trust as the biggest adoption barrier. The same research found that 60% lack a dedicated response plan and forensic tools for agent-driven fraud.
Those gaps change the payment infrastructure checklist. Agent payments need identity checks, clear spending permissions, real-time anomaly detection, audit trails, and dispute processes that account for actions taken by software on a user’s behalf.
China’s agent ecosystem is moving quickly. Alibaba’s Qwen app surpassed 10 million downloads in its first week, giving Alibaba another consumer entry point for AI-assisted tasks. Alipay’s payment tools add a transaction layer around that kind of agent activity.
The risk is not theoretical. AI agents can act quickly and make mistakes quickly, as one recent report about an AI agent deleting a company database showed. Payments teams cannot treat agent actions like ordinary human checkout behavior.
For enterprises, Alipay’s launch is a reason to review authorization rules now. Before AI agents get broader spending authority, teams should decide which transactions require human approval, how agent activity is logged, and who is responsible when an automated purchase is disputed.
Also read: Nvidia’s GTC 2026 keynote framed agentic AI as part of a larger shift in how businesses build and run AI infrastructure.
Read the full article here
