JPMorgan and Hyundai Card finished a proof-of-concept in which a Hyundai Card AI agent initiated a simulated foreign-exchange trade to maintain euro liquidity in a Kinexys blockchain deposit account.
The banks now target the first half of 2027 for a more complex phase where multiple AI agents could interact, negotiate and execute payments without direct human intervention.
JPMorgan said the harder challenge was governance, not coding, so Kinexys built controls that confine agents to preset limits — blocking, for example, transactions above a $5 million authorization cap.
The test was described as the first private blockchain-based autonomous bank-agent-to-client-agent trial for corporate treasury, though JPMorgan has not set a commercialization timeline.
The project extends Kinexys's broader push in blockchain payments, a business that had processed more than $4 trillion in transactions and averaged about $7 billion a day as of August.
Will JPMorgan's private blockchain successfully merge institutional safeguards with machine-speed settlement, or will rogue AI agents outsmart their own limits?
As JPMorgan prepares for AI agents to negotiate by 2027, could autonomous machine-to-machine payments trigger an unstoppable financial flash crash?
If an AI agent negotiates a multi-million dollar trade autonomously, who takes the fall when the algorithm makes a catastrophic error?