The products were unveiled on 26 May at Alipay’s AI Payment Ecosystem Conference in Hangzhou. The company presented them as part of a wider “full-stack” AI payment infrastructure aimed at consumers, AI model developers and retailers, with the goal of making payments easier to embed inside automated shopping and service agents.
AI Wallet is designed as a consumer-facing feature inside the Alipay app. Users can search for “AI Wallet” and manage tasks carried out by AI agents before and during payment, then review spending afterwards. The product is intended to address one of the most sensitive issues in agentic commerce: how far a consumer can delegate purchase decisions without losing visibility over authorisation, limits and accountability.
Token Pay is aimed at AI model companies, giving them tools for global subscription payments, token top-ups inside AI agents and related transaction needs. MiniMax and Stepfun have joined Alipay in new collaborations covering token top-ups, membership subscriptions and marketing campaigns, placing the launch within China’s wider contest to monetise large-model services and agent-based applications.
The move builds on Alipay AI Pay, which lets consumers complete transactions through AI agents, including by voice command. Alipay says the service had crossed 100 million users in February and has handled 300 million transactions to date, spanning retail apps, mini programmes, AI smart glasses, smart cockpits, development platforms and consumer-facing AI services such as Alibaba’s Qwen.
Cyril Han, chief executive of Ant Group, framed the launch as a shift in how commerce will operate as agents move from search and recommendation into execution. The company’s message is that payments can no longer be treated merely as the last step of a purchase; they must be built into the architecture of automated services from the outset.
That argument is gaining force across the global payments industry. Visa, Mastercard, Stripe, OpenAI, Google and specialist payment infrastructure firms are all developing protocols or standards to verify agents, authenticate user intent and tokenise transactions. The common aim is to let merchants distinguish approved agents from malicious bots while ensuring users retain control over spending.
Alipay’s trust framework reflects similar concerns. Its Agentic Commerce Trust Protocol is intended to create a common language between AI platforms and service providers, while its security system is meant to place safeguards around agent-led transactions. The company is also using Alipay+ infrastructure to promote agent-ready wallets, including spending limits, passkey-based authentication, intent governance, one-time payment credentials and agent registration.
The commercial prize is substantial. AI agents are being designed to compare prices, order food, book services, renew subscriptions and make low-value purchases without repeated manual checkout. For retailers, that could reduce friction and open new channels for sales. For AI model companies, embedded payments can help convert usage into revenue through subscriptions, consumption-based pricing and token top-ups.
The risks are also rising. Agentic commerce changes the fraud surface because legitimate agents may complete many transactions quickly, making old signals such as browsing behaviour, device fingerprints and timing less reliable. Fraud can also emerge from hijacked agents, excessive delegated permissions, fake merchant listings or purchases that technically follow a user’s broad instruction but fail to match the user’s intent.
Merchants face another challenge: accountability. If an AI agent buys the wrong item, overpays, selects a fraudulent merchant or misreads a user’s instruction, the dispute may involve the consumer, the wallet provider, the merchant, the AI platform and the payment network. That makes clear consent records, transaction logs, refund rules and dispute-handling standards essential.
Alipay’s scale gives it a strong starting point in China. The platform connects more than one billion consumers with more than 80 million merchants and integrates more than 10,000 daily-life services, including travel, healthcare, tourism and entertainment. That breadth gives Ant Group a large testing ground for agentic payments across both online and offline commerce.
Competition, however, is intensifying. Stripe and OpenAI have built an Agentic Commerce Protocol for purchases inside ChatGPT, while Visa and Mastercard are pursuing agent authentication and tokenised payment frameworks. These initiatives point to a fragmented but fast-forming market in which interoperability, merchant control and consumer trust may decide which systems gain adoption.
Topics
Technology