No Human Intervention Needed! JPMorgan (JPM.US) Teams Up with Hyundai Card to Explore Autonomous Financial Transactions by AI Agents
JPMorgan and Hyundai Card of South Korea 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.
Artificial intelligence (AI) agents have proven they can play a positive role, cause a lot of trouble, or even both at once. One application area seen as having enormous potential is payments. If sufficient trust can be established, AI agents could handle various money-related matters on people's behalf. That is why JPMorgan (JPM.US) and South Korea's Hyundai Card are testing how AI agents can handle financial transactions without direct human intervention, including whether they can one day negotiate with each other and complete payments.
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 involving multiple AI agents in the first half of 2027.
Wee Kee Toh, global head of Kinexys Insights at JPMorgan's blockchain subsidiary Kinexys, said the next step is to test what happens when autonomous agents move beyond transaction execution and begin interacting with each other within an ecosystem. "What happens when agents start negotiating with each other, reaching agreements directly, and then making payments? What would that process actually look like?" he said. "I think this is actually the most important part we want to further expand beyond the scope of the first phase of testing."
In the initial test, Hyundai Card's AI agent initiated a foreign exchange transaction request aimed at maintaining euro liquidity in its Kinexys blockchain deposit account. JPMorgan said this was the first corporate treasury management test on a private blockchain in which an autonomous bank agent interacted with a client agent.
Accountability Issues
Although the initial test has been completed, JPMorgan has not yet set a timetable for commercializing the technology.
Wee Kee Toh said building the technology itself is relatively simple, but establishing appropriate governance mechanisms and internal controls is more difficult. "From a technical perspective, this is actually not the biggest obstacle," he said, noting that what really needs to be resolved is the question of accountability and the risks of allowing AI agents to execute financial transactions without direct human oversight.
Experts say such agents have the potential to transform how interactions work in financial services, and that this transformation must come with corresponding accountability mechanisms. Data management and protection software company AvePoint found in its "2026 State of AI Report" that unauthorized autonomous agent behavior was the third most common type of security vulnerability related to AI agents, accounting for 34.1% among enterprises that reported such incidents. These behaviors include executing transactions and modifying data without approval, although the report's findings are not limited to corporate treasury payments.
"When agents negotiate and execute operations autonomously, the focus of governance shifts from monitoring human behavior to monitoring machine-to-machine dependencies," said Tianyi Jiang, CEO of AvePoint. He explained that interactions between AI agents can amplify errors and trigger transactions before humans have time to intervene.
"The worst-case scenario is a cascading failure. A compromised or hallucinating agent passes erroneous data to another agent," he said, adding that this could lead to "erroneous financial transfers occurring in rapid succession, triggering a chain reaction before humans even realize a problem is occurring."
To address these concerns, Kinexys has developed a framework that restricts AI agent activity to pre-set transaction parameters and automatically rejects requests exceeding authorized limits. "You need to set a boundary for the agent. The agent can never do anything beyond that boundary," Wee Kee Toh said. For example, if an AI agent is authorized to execute transactions up to US$5 million, any transaction exceeding that threshold would be blocked by the system.
Hassan Ahmed, head of Coinbase (COIN.US) Singapore, said that ensuring the security of agent payments requires clear safeguards, including spending limits, restricted account permissions, and explicit authorization requirements. He said that given AI could create new opportunities for malicious actors, users should also be able to monitor related activity at any time and revoke access at any time.
"While private or permissioned environments may offer additional controls in certain use cases, we should not overlook 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 JPMorgan and Hyundai Card are also exploring whether to expand their cooperation to other businesses under Hyundai Motor Group. Hyundai Motor Group's broad business scope is one reason JPMorgan chose Hyundai Card as a partner. No specific projects have been announced yet.
According to information on Hyundai Card's website, Hyundai Motor Group holds a combined 78.04% stake in Hyundai Card through Hyundai Motor, Kia, and Hyundai Commercial.
Hyundai Card has previously experimented with blockchain-based cross-border payments. In July this year, the company said it had completed a proof-of-concept test that transferred US$20,000 between Hyundai Motor's U.S. and Mexican subsidiaries using the USDT stablecoin, with the entire transfer and verification process taking an average of seven minutes.
Kinexys is also continuing to expand its blockchain-based payments business. The JPMorgan unit said in August this year that it had cumulatively processed more than US$4 trillion in transactions since inception, with average daily transaction volume of about US$7 billion. In addition, JPMorgan launched a proof of concept in June 2025 for its U.S. dollar-denominated deposit token JPMD and began offering the product to institutional clients in November of the same year.
Related Articles

Self-defined innovation sets a new paradigm for battery safety: Aurora Battery highlights BAIC MOTOR (01958) delivering a new "answer" for family-oriented all-terrain driving.
.png)
HK Stock Market Move | |JW THERAP-B (02126) Rebounds Over 8%; Company Clarifies Cost Dispute, Expects Unit Production Cost of Benlysta to Drop Over 65% Starting Next Year

HK Stock Market Move | LENOVO GROUP(00992) rises over 4%; Nvidia's earnings report confirms Lenovo AI PCs have shipped; the company is a launch partner for Vera Rubin.
Self-defined innovation sets a new paradigm for battery safety: Aurora Battery highlights BAIC MOTOR (01958) delivering a new "answer" for family-oriented all-terrain driving.

HK Stock Market Move | |JW THERAP-B (02126) Rebounds Over 8%; Company Clarifies Cost Dispute, Expects Unit Production Cost of Benlysta to Drop Over 65% Starting Next Year
.png)
HK Stock Market Move | LENOVO GROUP(00992) rises over 4%; Nvidia's earnings report confirms Lenovo AI PCs have shipped; the company is a launch partner for Vera Rubin.






