With the recent development of artificial intelligence (AI) agents, structural changes in the global..

사진 확대

Introduction screen to announce the launch of the “x402 Foundation” involving Coinbase and Cloudflare. [Source = x402 Foundation]

With the recent development of artificial intelligence (AI) agents, structural changes in the global payment and settlement market are beginning.

Beyond simply changing payment methods, the machine-to-machine (M2M) payment structure, in which AI agents perform payments on their own, is quickly being established as a payment infrastructure.

According to the IT and virtual asset industries on the 13th, competition for global standards to lead the AI agent-based payment ecosystem is intensifying.

The “x402 Foundation,” which was officially launched under the Linux Foundation on the 2nd, involved a large number of major big tech and payment operators such as Google, Microsoft, Amazon Web Services (AWS), Visa, and Mastercard.

In particular, in Korea, Kakao Pay joined as the only founding member, and has entered the establishment of AI-led global standards for “agent commerce.”

◆ A ‘three-layer’ structure of authorization, transaction creation, and payment execution

In this regard, the Yulchon Digital Asset Center, a law firm, recently published a report that analyzed that this AI agent payment infrastructure consists of a three-layer structure in which △ authority setting △ transaction creation △ payment execution is separated.

First, in the ‘authority layer’, users set the scope and limits of transactions to AI agents through AP2 (Agent Payment Protocol), etc., granting them unstructured and dynamic delegation authority.

After that, in the ‘transaction layer’, AI agents search and compare information in real time to create autonomous transactions that have not been confirmed in advance.

In the last “payment layer,” when the server sends an HTTP 402 code that says it needs proof of payment, the agent immediately uses the previously approved stablecoins of the wallet to settle the account.

All of these processes take place in milliseconds without human intervention, and realize ultra-small and high-frequency payments.

◆ Conflict with old financial regulations…Controversy over who is responsible for an accident

사진 확대

A technical introduction screen of the x402 protocol that supports Internet-native payments with a single line of code and utilizes HTTP 402 responses. [Source = x402 Foundation]

The Yulchon Digital Asset Center points out that the issue of consistency with the existing legal system is emerging as a key issue as ultra-small and real-time AI agent payments begin in earnest.

It is pointed out that the current Electronic Financial Transactions Act, designed on the premise of human instructions and intermediary intervention, is difficult to regulate AI automatic payments made in units of API calls without user intervention.

Above all, it is considered a practical challenge to whom and how to attribute the responsibility when an AI agent, a software that does not have an independent legal subjectivity, trades or makes an error beyond its authority.

In addition, if the autonomous price optimization process leads to collusion, preemptive research is needed to see if it can be regulated by algorithmic collusion under the Fair Trade Act.

Anti-money laundering (AML) schemes should also be redesigned as an integrated management of account or agent units rather than individual transaction units.

In addition, if the won-based stablecoin infrastructure is insufficient, the dollar-based payment structure is likely to become the market standard, so it is urgent to establish a strategy in terms of foreign exchange regulation.

◆ “We need to secure technical defense measures such as log data for the entire process.”

Experts are ordering preemptive preparations from companies to minimize risks that may arise in the AI agent payment market in the future. It is necessary to quickly establish a structure for sharing responsibilities between service providers, platforms, and payment infrastructure.

Kim Ik-hyun, head of the Yulchon Digital Asset Center at the law firm, said, “In AI agent payments, functions are carried out by different subjects, so it will be necessary to switch from existing business-oriented regulations to function-oriented regulations. Companies must come up with technical measures to record and preserve the entire process from authorization (AP2) to actual execution (x402) as log data in a form that cannot be falsified so that they can use it as a legal defense in the event of a dispute.”