No human intervention is required! Xiaomo (JPM.US) teamed up with Hyundai Credit Card to explore autonomous financial transactions with AI agents

Zhitongcaijing · 3d ago

The Zhitong Finance App learned that artificial intelligence (AI) agents have proven that they can play a positive role, cause a lot of trouble, or even both. One application area that is considered to have great potential is the payment business. If sufficient trust can be established, AI agents can handle all kinds of finance-related matters on behalf of people. Because of this, JPMorgan Chase (JPM.US) and Korea's Hyundai Card Company (Hyundai Card) are testing how AI agents can handle financial transactions without direct human intervention, including whether they can negotiate with each other and complete payments in the future.

The two companies have completed an initial proof of concept involving a simulated foreign exchange transaction, and plan to launch a more complex testing phase in the first half of 2027, which will involve multiple AI agents.

Wee Kee Toh, global head of Kinexys Insights, a blockchain subsidiary of J.P. Morgan Chase, said the next step is to test what happens after autonomous agents go beyond transaction execution and begin to interact with each other in an ecosystem. He said, “What happens when smart entities begin to negotiate with each other, directly reach an agreement, and then make payments? What would this process actually look like?” “I think this is actually the most important part we want to expand further on the scope of the first phase of testing.”

According to reports, in the initial test, Hyundai Credit Card's AI agent initiated a foreign exchange transaction request with the aim of maintaining the euro liquidity in its Kinexys blockchain deposit account. According to J.P. Morgan Chase, this is the first enterprise fund management test based on a private blockchain where autonomous banking agents interact with customer intelligence.

Accountability issues

Although initial testing has been completed, J.P. Morgan has yet to set a timeline for commercializing this technology.

Wee Kee Toh said the technology itself is relatively simple to build, but establishing proper governance mechanisms and internal control measures is more difficult. “From a technical point of view, this isn't actually the biggest obstacle,” he said. He pointed out that what really needs to be solved is the problem of attribution of responsibility and the risks posed by allowing AI agents to execute financial transactions without direct human supervision.

Experts say this type of intelligence has the potential to change the way financial services interact, and this transformation must be accompanied by corresponding accountability mechanisms. Data management and protection software company AvePoint found in its “2026 AI Status Report” that unauthorized autonomous agent behavior is the third most common type of security vulnerability related to AI agents, accounting for 34.1% of companies reporting such incidents. These acts include execution of transactions and data modification without approval; however, the findings of the report are not limited to the field of corporate money management payments.

Tianyi Jiang, CEO of AvePoint, said: “When agents independently negotiate and execute operations, the focus of governance efforts will shift from monitoring human behavior to monitoring the dependency relationships between machines.” He explained that interaction between AI agents may amplify errors and trigger transactions before humans have time to intervene.

He said, “The worst case scenario is a chain failure. An invaded or illusory agent passes erroneous data to another agent.” He added that this could cause “erroneous financial transfers to occur one after another at a very rapid pace, triggering a cascade of reactions even before humans realized the problem was occurring.”

To address these concerns, Kinexys has developed a framework that limits AI agent activity to pre-set transaction parameters and automatically rejects requests that exceed authorized limits. Wee Kee Toh said, “You need to draw a boundary for the smart body. An intelligent body can never do anything beyond this boundary.” As an example, he said that if an AI agent is allowed to execute a maximum transaction amount of 5 million US dollars, then any transaction that exceeds this threshold will be blocked by the system.

Hassan Ahmed, head of Coinbase (COIN.US) Singapore, said that to ensure the security of smart payments, it is necessary to establish clear protective measures, including setting spending limits, limiting account permissions, and clarifying authorization requirements. He said that considering that AI may create new opportunities for malicious actors, users should also be able to monitor related activities at any time and revoke access rights at any time.

“Although private or permissioned environments may provide additional controls in some application scenarios, we should not lose sight of the potential of open blockchain networks to support a more inclusive, interoperable, and efficient financial system,” Hassan Ahmed said.

Beyond the pilot phase

Wee Kee Toh said that J.P. Morgan Chase and Hyundai Credit Card are still exploring the possibility of expanding their cooperation to other businesses under the Hyundai Motor Group. The Hyundai Motor Group has a wide range of businesses, which is one reason why J.P. Morgan Chase chose Hyundai Credit Card as its partner. Currently, the two sides have not announced any specific projects.

According to Hyundai Credit Card's website, Hyundai Motor Group holds a total of 78.04% of Hyundai Credit Card's shares through Hyundai Motor Company, Kia, and Hyundai Commercial.

Previously, modern credit card companies had tried blockchain-based cross-border payments. In July of this year, the company said it had completed a proof-of-concept test and transferred 20,000 US dollars between Hyundai Motor's US subsidiary and Mexican subsidiary through USDT stablecoins. The entire transfer and verification process took an average of 7 minutes.

Kinexys is also continuing to expand its blockchain-based payments business. The business unit under J.P. Morgan Chase said in August this year that since its inception, the cumulative transaction amount it has processed has exceeded 4 trillion US dollars, with an average daily transaction volume of about 7 billion US dollars. Additionally, J.P. Morgan Chase launched a proof of concept for its dollar-denominated deposit token JPMD in June 2025, and began providing the product to institutional customers in November of the same year.