How can industry giants secure AI payment systems while vying for market share?

Who will subvert the "traffic is king" logic in the mobile Internet era?

Han Xinyi, CEO of Ant Group, gave his judgment - AI Agent.

On May 26th, Han Xinyi put forward a brand - new view at the Alipay AI Payment Ecosystem Conference: In the future, the era of "traffic entrances" determined by humans is coming to an end, and it will be replaced by the "ecosystem of agents".

The technological change behind this is that with the rapid evolution of large AI models, today's AI "agents" can not only chat but also have the ability to independently identify user needs, plan tasks, and even place orders and make payments directly.

In the past, planning a trip meant making your own travel guides, booking flights, hotels, and ride - hailing services. In the future, you may only need to say to the agent, "Plan a perfect trip."

In this process, AI is gradually transforming from an auxiliary tool into a "subject" that can directly participate in economic activities. This is a leap in the agent's capabilities and also foreshadows a complete reconstruction of the existing business logic.

Since 2025, AI payment based on agents has become a battleground for technology giants and payment behemoths.

Domestically, institutions such as WeChat Pay (Tenpay), JD Technology, and China UnionPay have laid out the AI payment ecosystem by releasing AI payment skills, agent payment protocol frameworks, etc.

Internationally, institutions such as Google, VISA, Mastercard, Stripe, and PayPal have also carried out ecological positioning through agent payment protocols, bridging web3 and cryptocurrency payments, etc.

In Han Xinyi's view, in the upcoming agent era, whoever can closely connect users, merchants, AI model companies, developers, and their agents to build a huge ecosystem will have a more solid business moat. And Alipay is trying to build a new ecosystem for agent - based business.

At the conference, Alipay announced that the platform has achieved over 300 million payments based on AI agents. Li Jiajia, co - president of Ant Group's Digital Payment Business Group, said in an interview with media such as Caixin: "The number increased by 200 million in the past two months, and it may increase by 1 billion in the next two months. We believe it will come soon, and we hope to speed up the process."

Zhu Min, former deputy governor of the People's Bank of China, said in a recent interview with Caixin that since the beginning of the year, the combination of payment and agents has brought the biggest change to the global financial industry.

Meanwhile, agents are shifting payment decision - making from "human - made decisions" to "joint decisions made by humans and agents". In the future scenarios envisioned by many industry insiders, payment may even become "autonomous decision - making by agents" to some extent.

In response, Zhu Min warned that if financial agents achieve autonomous decision - making and autonomous payment, it may bring risks to a series of fields such as privacy protection, decision - making responsibility determination, legal compliance, and cross - border supervision, which requires forward - looking supervision to address.

Tech Giants Compete for AI Payment

As early as September 2025, Alipay first announced the launch of the "AI Pay" service, providing payment support for agents in the AI era. It was first launched on Luckin Coffee's AI ordering assistant "Lucky AI", allowing users to complete orders and payments directly within the agent through natural language interaction of "talking".

In April 2026, Alipay launched the "AI Receive" function for merchants and individual developers providing AI services, saving developers the trouble of building a complex payment and settlement system.

Currently, Alipay's AI payment supports 95% of general agent frameworks. This includes general agents such as Qianwen, JVS Claw, Claude code, and Hermes Agent; smart devices such as Qianwen AI glasses, Rokid, and future smart AI earphones; smart cockpits of Ideal Auto, Chery, Geely, Dongfeng, etc.; and leading AI tool platforms such as TRAE SOLO, Qoder, and Kouzi.

In fact, technology and payment giants at home and abroad are competing for the AI payment ecosystem.

On May 15th, JD Group announced a strategic cooperation agreement with the international card organization Mastercard. JD said that the two sides will explore providing AI - agent payment services for individuals and enterprises through Mastercard's Agent Pay intelligent payment solution.

Xu Ran, vice - chairperson of JD Group's SEC and CEO of JD Group, said that JD will jointly explore the next - generation business and payment experience driven by advanced technology and artificial intelligence with Mastercard.

Even earlier, technology giants and payment behemoths had already started forming alliances.

China UnionPay released the Agentic Payment Open Protocol (APOP), aiming to build a new, open, controllable, secure, and trustworthy agent payment ecosystem. According to information on UnionPay's official website, a total of 19 domestic and foreign institutions became APOP partners in the first batch, including commercial banks, agent manufacturers, technology companies, merchants, and acquiring institutions.

At the beginning of 2026, Alipay, in cooperation with Qianwen App, Taobao Flash Sale, Damai, and Alibaba Cloud Bailian, released the Agentic Commerce Trust Protocol (ACT), aiming to solve the security and trust issues of AI agents during independent operation, cross - platform collaboration, and proxy payment for customers.

Even earlier, in September 2025, Google released the Agent Payments Protocol (AP2), providing a native and intermediary - free payment method for interactions between agents. Google said that the partners of this protocol include more than 60 global organizations such as Mastercard, UnionPay International, Ant International, PayPal, and Coinbase, covering payment giants, financial institutions, and technology companies.

Since 2026, Google has further introduced the Verifiable Intent trust layer and the Universal Commerce Protocol (UCP) to further improve the agent - proxy payment specifications from payment authorization to the entire shopping transaction process.

An industry insider in AI payment told Caixin that these protocols can be regarded as industry standards or norms for agent - proxy payment, and standards often mean the right to speak. In the early stages of the development of industries such as credit cards and 5G, there were often a variety of technical standards. As competition converges, various standards begin to interconnect.

"AI payment is currently in the stage of a 'diverse range of standards'," said the aforementioned industry insider.

It is worth noting that Zhu Min said in an interview with Caixin that when thousands of agents exist in the same payment system, the scale of funds affected will be beyond imagination. If agents in a unified system make a unified decision in an instant, a "herd effect" will occur. The combination of the two may trigger a "tsunami - level" systemic risk.

"The 'tsunami' process triggered by financial agents will cross numerous regulatory boundaries, and there is still a lack of consensus on regulation in this field at present," Zhu Min warned. The technological iteration, intelligence, speed, and scale of financial agents will far exceed the existing regulatory system, and a series of such changes require forward - looking supervision.

How to Build a Security Barrier

In the agent era, scale and network effects still determine the business moat.

Han Xinyi said in his opening speech that in the mobile Internet era, "traffic is king", and the access traffic generated by humans is the business moat. In the agent era, the "ecosystem of agents" will be a larger moat, and the scale of agents and the network effects between agents are very crucial.

Li Jiajia told Caixin that currently, Alipay does not consider business, revenue, or competition in its AI payment business. It only wants to play a good role as the AI payment infrastructure and first expand the ecosystem. "When the ecosystem expands, business revenue will naturally follow," Li Jiajia said. One of the current phased key performance indicators for Alipay's AI payment business is the number of agent developers on the agent platforms within the ecosystem.

In fact, this is also the consensus among players in the current AI payment track.

In February 2026, the cumulative number of weekly payment transactions and the number of users of Alipay's "AI Pay" exceeded 100 million respectively. In the following two months, the number of AI payment transactions continued to rise and has now exceeded 300 million.

Although the number of agent payment transactions is rising rapidly, the entire industry ecosystem is still in its early stage. Li Jiajia revealed in an interview with the media that at least 90% of the 300 million AI agent payments come from Qianwen, a general agent under Alibaba. "Currently, the scenarios where AI payment is accessed all require developers, and the scenario that most needs AI payment in China is Qianwen," Li Jiajia said.

At the conference, Alipay launched the Token Pay service to solve the model payment solution for large - model manufacturers, allowing them to focus on model development itself. At the same time, Alipay announced the "Alipay AI Developer Incentive Program" on - site, encouraging more people to develop agents with charging behaviors through measures such as token computing power subsidies, payment rate discounts, and payment scenario provision.

"But before that, another problem needs to be solved: How can consumers dare to use AI payment?" Li Jiajia admitted that when Qianwen first launched the AI payment function, many customers' feedback was that they "didn't dare to use it".

To this end, Alipay launched the "AI Wallet" for individual users. Zhu Lin, the general manager of Ant Group's AI payment, said that the "AI Wallet" is designed to solve the problem of agent management and make users dare to use AI payment.

Specifically, the "AI Wallet" can enable users to control the entire process of agent payment through pre - payment intention authorization management, 24 - hour task execution supervision and risk warning during payment, and intelligent consumption management of AI bills after payment.

In addition, at the aforementioned event, Alipay included AI payment in the "If you dare to pay, I dare to compensate" commitment to cover the risks of AI payment.

However, the changes brought by AI payment go far beyond this. If consumers and merchants conduct transactions through their respective agents, an Agent - to - Agent (A2A) payment model between agents is created. In this model, who can be responsible for the agents' actions?

An industry insider familiar with the relevant situation told Caixin that A2A payment is also a scenario that regulators are very concerned about, including how to confirm the identity of agents and how to handle the authorization of agents during the A2A payment process.

In fact, a financial regulatory official emphasized at a round - table forum recently that the application of AI technology in the financial industry still needs to achieve "human - machine collaboration". "We should give full play to human advantages in value judgment, ethical decision - making, innovation, and creation. We should ensure that humans make the final decisions, supervise key links, and intervene in emergency situations. This is the current state that human - machine collaboration should be in."

This article is from the WeChat official account "Caijing May Flower" (ID: Caijing - MayFlower). Author: Tang Jun, Editor: Zhang Wei. Republished by 36Kr with permission.