JPMorgan and Hyundai Card test autonomous AI agents executing financial transactions
JPMorgan and Hyundai Card are testing autonomous AI agents that can execute financial transactions independently, marking a shift toward machines managing money without real-time human approval. The challenge now is not technical but regulatory and operational: building governance frameworks that let AI agents act within strict boundaries while managing the accountability risks of unsupervised financial decision-making.
- JPMorgan and Hyundai Card completed a proof-of-concept where an AI agent autonomously requested a foreign exchange transaction to maintain euro liquidity.
- The companies are targeting the first half of 2027 to test multiple AI agents negotiating and transacting with one another in an ecosystem.
- Kinexys has developed a framework that restricts AI agents to predefined transaction parameters, automatically rejecting requests outside authorized limits.
- $4 trillion Total transactions processed by Kinexys since inception, averaging $7 billion daily
- $5 million Example transaction ceiling threshold beyond which an AI agent would be automatically blocked
- 7 minutes Average time for $20,000 USDT stablecoin transfer between Hyundai subsidiaries in July 2026
- November 2025 Launch date of JPM Coin to institutional clients after June 2025 pilot
According to CNBC reporting, JPMorgan Chase and South Korea’s Hyundai Card are pioneering a system in which artificial intelligence agents can independently manage financial transactions on behalf of organizations, without requiring human authorization for individual deals. The initial test involved a Hyundai Card AI agent autonomously requesting a foreign exchange transaction to maintain euro liquidity in its account on Kinexys, JPMorgan’s blockchain platform. Kinexys has processed more than $4 trillion in transactions since inception, averaging about $7 billion a day, and the companies now plan to escalate their trials significantly in coming months.
AI agents negotiating directly with each other in 2027
Wee Kee Toh, global head of Kinexys Insights at Kinexys by JPMorgan, told CNBC that the next phase moves beyond simple execution to agent-to-agent interaction. “That, I would say, is really the biggest part of what we want to get as an expansion of the phase one scope,” Toh said when describing plans for the first half of 2027, when the companies will test whether multiple autonomous agents can interact with and transact with one another in a shared financial ecosystem.
Technology-wise, it’s really not the biggest hurdle. You are going to draw a box around the agent. The agent will never be able to do anything out of the box.
Wee Kee Toh, global head of Kinexys Insights at Kinexys by JPMorgan, told CNBC
JPMorgan described the initial trial as the first private blockchain-based autonomous bank-agent-to-client-agent test for corporate treasury management.
Governance and control constraints replace technical complexity
Toh identified the core obstacle to deploying autonomous financial agents: not engineering but risk management. The technical infrastructure for AI agents to execute transactions proved straightforward, but designing appropriate governance frameworks and internal controls to prevent misuse required substantially more work.
To address those concerns, Kinexys developed a framework that restricts AI agents to predefined transaction parameters. An AI agent authorized to transact up to $5 million would be automatically blocked from executing transactions above that ceiling, Toh explained. The system rejects requests that fall outside authorized limits before they can be processed, ensuring agents remain bound by pre-set boundaries regardless of what instructions they receive.
JPMorgan has not yet set a timeline for bringing this technology to market at scale.
Hyundai Card’s blockchain track record and future partnership scope
Hyundai Card was not new to blockchain experimentation when JPMorgan selected it as a partner. In July 2026, the company said it completed a proof-of-concept involving a $20,000 transfer using the USDT stablecoin between Hyundai Motor’s U.S. and Mexican subsidiaries, with the transfer and verification process taking an average of seven minutes compared to traditional cross-border payment timelines measured in days.
JPMorgan selected Hyundai Card partly because of the broader operations within Hyundai Motor Group, which holds a combined 78.04% stake in Hyundai Card through Hyundai Motor, Kia and Hyundai Commercial. The companies are exploring whether to extend their collaboration to other businesses within the group, though Toh said no specific projects have been announced yet.
The BlockWest read. The real shift here is structural, not technological. JPMorgan’s existing blockchain infrastructure at Kinexys can already handle the volume and speed; what matters is that this partnership tests whether financial institutions trust machines to commit capital autonomously within guardrails. That trust—and the governance models that earn it—will determine whether AI agents become a standard tool for treasury departments or remain a curiosity.
The timing to watch is the first half of 2027, when JPMorgan and Hyundai Card plan to test multiple agents interacting directly with each other. That milestone will show whether the governance framework holds under real-world conditions where agents must negotiate with counterparties whose behavior is not fully known in advance. JPMorgan has not committed to a commercialization date, leaving the question of when institutions can actually deploy this capability for their own operations still open.
BlockWest is a news publication. Nothing here is investment advice. Read our disclaimer and editorial policy.
