BTCBTC$61,544-3.37%|
ETHETH$1,735.85-3.70%|
USDTUSDT$0.99922-0.00%|
BNBBNB$566.8800-4.12%|
USDCUSDC$0.99966+0.00%|
XRPXRP$0.94441-4.29%|
SOLSOL$70.9400-4.98%|
TRXTRX$0.30561-2.22%|
FIGR_HELOCFIGR_HELOC$1.0160-0.65%|
HYPEHYPE$62.9500-8.25%|
ZECZEC$809.9100-14.11%|
DOGEDOGE$0.05645-5.19%|
XMRXMR$461.9800-8.19%|
LEOLEO$9.2300-1.14%|
WBTWBT$59.6300-3.42%|
LINKLINK$8.7400-5.13%|
USDSUSDS$0.99986-0.02%|
ADAADA$0.14620-5.83%|
USDEUSDE$0.99927-0.02%|
XLMXLM$0.13132-4.75%|
DAIDAI$0.99988+0.01%|
BCHBCH$196.4500-6.06%|
NEARNEAR$2.9600-12.69%|
LTCLTC$41.4000-6.75%|
CCCC$0.07869-5.68%|
AVAXAVAX$6.7900-4.72%|
GRAMGRAM$1.0130-1.57%|
USD1USD1$0.99891-0.03%|
XAUTXAUT$3,563.29-3.86%|
UNIUNI$4.7300-7.85%|
BTCBTC$61,544-3.37%|
ETHETH$1,735.85-3.70%|
USDTUSDT$0.99922-0.00%|
BNBBNB$566.8800-4.12%|
USDCUSDC$0.99966+0.00%|
XRPXRP$0.94441-4.29%|
SOLSOL$70.9400-4.98%|
TRXTRX$0.30561-2.22%|
FIGR_HELOCFIGR_HELOC$1.0160-0.65%|
HYPEHYPE$62.9500-8.25%|
ZECZEC$809.9100-14.11%|
DOGEDOGE$0.05645-5.19%|
XMRXMR$461.9800-8.19%|
LEOLEO$9.2300-1.14%|
WBTWBT$59.6300-3.42%|
LINKLINK$8.7400-5.13%|
USDSUSDS$0.99986-0.02%|
ADAADA$0.14620-5.83%|
USDEUSDE$0.99927-0.02%|
XLMXLM$0.13132-4.75%|
DAIDAI$0.99988+0.01%|
BCHBCH$196.4500-6.06%|
NEARNEAR$2.9600-12.69%|
LTCLTC$41.4000-6.75%|
CCCC$0.07869-5.68%|
AVAXAVAX$6.7900-4.72%|
GRAMGRAM$1.0130-1.57%|
USD1USD1$0.99891-0.03%|
XAUTXAUT$3,563.29-3.86%|
UNIUNI$4.7300-7.85%|
News/AI & Blockchain

JPMorgan and Hyundai Card Test AI Agents for Autonomous Currency Trading on Kinexys Blockchain

JPMorgan and Hyundai Card have completed a blockchain-based trial in which AI agents autonomously managed simulated foreign exchange transactions. Using Kinexys, the agents monitored balances, compared currency rates and executed authorized requests without direct human intervention. The companies plan more advanced agent-to-agent transaction testing in 2027.

Aaron Bailey
By Aaron Bailey
JPMorgan and Hyundai Card AI Agents Test Blockchain FX

Key Takeaways

  • Successful AI banking trial: JPMorgan and Hyundai Card completed the first phase of a proof-of-concept involving autonomous financial transactions.
  • Automated foreign exchange: AI agents monitored account balances, compared currency quotes and simulated converting British pounds into euros.
  • Kinexys blockchain infrastructure: The experiment used JPMorgan's permissioned blockchain banking platform.
  • No real money transferred: All transactions were simulated in a controlled test environment.
  • Built-in security controls: AI agents operated under digitally signed mandates, verified identities and predefined transaction limits.
  • Human oversight remained: Transactions outside authorized limits could be rejected or escalated to employees.
  • 2027 expansion: The companies are targeting more complex multi-agent negotiations and payment scenarios in the first half of 2027.

JPMorgan and Hyundai Card AI Agents Complete Blockchain-Based Currency Trading Trial

The JPMorgan and Hyundai Card AI agents experiment marks an important development in the emerging field of autonomous financial services.

On November 18, 2026, JPMorgan's blockchain division, Kinexys, and South Korean financial services company Hyundai Card announced that they had completed the first phase of a proof-of-concept designed to test whether artificial intelligence agents could independently perform corporate treasury operations.

The experiment demonstrated how AI software could monitor account balances, identify funding shortages, compare foreign exchange rates and initiate currency conversion requests without requiring a human employee to manually perform every step.

The test was conducted using Kinexys, JPMorgan's institutional blockchain platform.

However, the transactions were simulated, and no real funds moved between accounts.

This distinction is important because the experiment demonstrated the feasibility of autonomous transaction workflows rather than the launch of a commercially available AI banking service.

The initiative focused on one of the most important challenges facing AI-powered finance: allowing software agents to make financial decisions while ensuring that those decisions remain subject to institutional authorization and compliance controls.

Unlike conventional automation, where software follows a fixed sequence of instructions, the AI agents in the trial could evaluate financial conditions and determine appropriate actions within predefined boundaries.

JPMorgan and Hyundai Card now intend to explore more sophisticated scenarios involving multiple AI agents negotiating and executing transactions.

The next phase is expected to take place during the first half of 2027.

How JPMorgan and Hyundai Card AI Agents Executed Currency Transactions

The proof-of-concept focused on a realistic corporate treasury management scenario.

Large multinational companies often maintain bank accounts in different countries and currencies.

One subsidiary may temporarily hold excess cash while another requires additional funds to meet operational obligations.

Traditionally, treasury employees or automated banking systems monitor balances and arrange transfers when necessary.

JPMorgan and Hyundai Card tested whether AI agents could coordinate this process with less manual intervention.

Step 1: AI detects a liquidity shortage

The simulated scenario involved a Hyundai Card account in Frankfurt holding euros.

The AI agent was instructed to maintain the account above a predetermined minimum balance.

When the balance fell below that threshold, the agent identified the shortage.

Rather than waiting for an employee to review the account, the software began evaluating how to restore the required liquidity.

Step 2: AI compares foreign exchange quotes

The agent determined that British pounds held in a London account could be converted into euros to cover the shortage.

It then requested foreign exchange quotes from three banks.

The objective was to identify the most competitive available exchange rate.

This is significant because the AI agent was not merely executing a predetermined payment.

It was evaluating multiple available options before deciding which transaction to request.

Step 3: Hyundai Card's agent contacts Kinexys

After selecting a quote, Hyundai Card's AI agent submitted the proposed transaction to the Kinexys AI agent.

The request contained information about the intended currency conversion and the authority under which the agent was operating.

The receiving system did not automatically trust the request simply because another AI agent had generated it.

Instead, Kinexys applied its verification framework.

Step 4: Kinexys verifies identity and authorization

The Kinexys agent checked the requesting agent's identity and permissions.

It also evaluated whether the transaction complied with a digitally signed mandate approved by the responsible institution.

The mandate established which activities the agent could perform and the conditions under which it could act.

If the transaction satisfied the relevant requirements, the system could approve execution automatically.

If the request exceeded the authorized scope, the system could reject it or escalate it for human review.

Step 5: The simulated FX transaction is executed

After the request passed the required checks, the Kinexys agent automatically executed the authorized workflow and booked a simulated foreign exchange transaction.

The system also recorded the reasoning, authorization checks and transaction outcome in an auditable record.

This created a traceable history of the automated decision.

The experiment therefore demonstrated more than automated currency conversion.

It tested whether financial institutions could establish accountability for decisions initiated by autonomous software.

Why the JPMorgan and Hyundai Card AI Agents Trial Matters for Banking

Artificial intelligence is already widely used in financial services.

Banks employ AI systems for fraud detection, customer support, risk analysis and financial forecasting.

However, allowing AI software to independently initiate financial transactions introduces a different level of responsibility.

A chatbot that provides an incorrect answer may create inconvenience.

An AI agent that transfers money incorrectly could create direct financial losses.

For this reason, banks need mechanisms to determine who authorized an AI agent, what actions it can perform and how its decisions can be reviewed.

The JPMorgan and Hyundai Card experiment addressed these concerns through a framework combining digital identity, delegated authority and transaction controls.

The model could eventually support corporate treasury activities that currently require employees to monitor accounts and manually approve routine transactions.

For example, a multinational business might authorize an AI agent to maintain minimum liquidity across several subsidiaries.

The agent could monitor balances, evaluate available currencies and propose transfers when funding requirements change.

However, it would remain subject to rules established by the company and its banking partners.

This distinction between autonomy and unrestricted financial control is central to the project.

The objective is not to allow AI systems to move money without accountability.

It is to automate permitted financial activities while maintaining institutional oversight.

What Is Kinexys, JPMorgan's Blockchain Banking Platform?

Kinexys is JPMorgan's blockchain business unit and institutional digital asset infrastructure.

The platform was previously known as Onyx before being rebranded in November 2024.

It supports financial applications involving blockchain-based deposits, payments and asset tokenization.

Unlike public blockchain networks such as Ethereum or Solana, Kinexys operates within a permissioned institutional environment.

Participation is restricted to authorized entities.

This allows financial institutions to establish identity verification, compliance requirements and transaction controls.

The platform is designed to support financial operations that may benefit from programmable settlement and shared transaction records.

Its blockchain deposit infrastructure allows participating institutions to conduct eligible transactions using digital representations of bank deposits.

For the Hyundai Card experiment, Kinexys provided the institutional environment in which the simulated treasury workflow could be evaluated.

The AI agents were integrated with the platform's authorization and control systems.

This allowed the test to examine how autonomous software could interact with financial infrastructure without bypassing established banking responsibilities.

The initiative also illustrates how blockchain technology and artificial intelligence can perform complementary roles.

AI agents can evaluate information and recommend actions.

Blockchain-based infrastructure can provide transaction records and support enforcement of predefined execution rules.

However, blockchain technology alone does not guarantee that an AI agent's financial decisions are correct.

The quality of the decisions still depends on the underlying models, available information and institutional controls.

Why Digital Identity and Signed Mandates Are Essential for AI Payments

One of the most important findings from the experiment concerns the security framework.

When a human employee initiates a bank transfer, the institution can verify their identity, account permissions and transaction authority.

AI agents require equivalent safeguards.

JPMorgan's approach uses digitally signed mandates to establish the relationship between an AI agent and the institution authorizing its actions.

A mandate defines what an agent is permitted to do.

For example, a company might allow an agent to convert currencies only when a particular account falls below a minimum balance.

The mandate could also specify transaction limits, eligible accounts and conditions requiring human approval.

In the Kinexys framework, an agent is associated with a cryptographic identity linked to its institutional principal.

Its messages can be authenticated using digital signatures.

The system can then determine whether the requested transaction complies with the approved mandate.

This helps prevent an AI agent from independently expanding its own authority.

The framework also maintains an audit trail documenting the transaction process.

For financial institutions, these safeguards are critical.

Without them, autonomous financial systems could introduce risks involving unauthorized transfers, incorrect decisions and unclear responsibility.

JPMorgan's experiment demonstrates one possible architecture for addressing those concerns, although a successful proof-of-concept does not eliminate the need for further security and operational testing.

AI Agents Could Transform Corporate Treasury Management

The most immediate application for this technology is corporate treasury management.

Large multinational companies regularly move money between subsidiaries to manage liquidity, pay suppliers and maintain operating balances.

These processes can involve multiple currencies, banks and jurisdictions.

Foreign exchange rates may also change throughout the day.

AI agents could potentially monitor these conditions continuously and identify suitable transactions.

Consider a company operating in South Korea, Germany and the United Kingdom.

Its German subsidiary may need additional euros to cover operating expenses.

Meanwhile, its UK subsidiary may hold excess pounds.

An authorized AI agent could identify the funding requirement, compare available exchange rates and request a transfer through approved banking infrastructure.

If the transaction satisfies the company's rules, execution could proceed without a treasury employee manually approving every routine step.

This could reduce operational workload and improve responsiveness.

However, the actual benefits remain unquantified.

JPMorgan and Hyundai Card have not disclosed verified cost savings, settlement-speed improvements or transaction volumes from the experiment.

The potential efficiency gains therefore remain a business hypothesis requiring further testing.

Hyundai Card's Earlier Stablecoin Trial Provides Important Context

The Kinexys experiment follows Hyundai's broader exploration of blockchain-based payments.

In July 2026, Hyundai Card participated in a cross-border stablecoin remittance pilot involving Hyundai Motor's overseas operations.

The earlier experiment used Tether's USDT stablecoin on Avalanche to transfer approximately $20,000 between the company's US and Mexican subsidiaries.

The reported transfer was completed in approximately seven minutes.

That project differed from the new JPMorgan experiment in several ways.

The stablecoin trial examined the movement of actual digital value across borders.

The Kinexys proof-of-concept focused on autonomous decision-making, authorization and simulated foreign exchange transactions.

Together, the projects illustrate two potential components of future financial infrastructure.

Blockchain-based payment systems can provide digital settlement mechanisms.

AI agents could potentially determine when and how authorized payments should occur.

However, the two technologies do not necessarily need to be combined in every financial application.

Traditional banking APIs can also support automated payments.

The commercial value of integrating AI agents with blockchain infrastructure will depend on whether the combination delivers measurable advantages over existing systems.

JPMorgan's AI Experiment Reflects a Wider Banking Tokenization Trend

The JPMorgan and Hyundai Card collaboration is part of a larger movement toward programmable financial infrastructure.

Banks are exploring how blockchain-based deposits, tokenized securities and automated settlement systems can work together.

Singapore has recently emerged as an important testing ground for these developments.

On November 19, the Monetary Authority of Singapore announced the successful issuance of S$100 million in tokenized MAS Bills settled using wholesale central bank digital currency.

That initiative demonstrated how digital central bank money could be used alongside tokenized securities in an institutional environment.

Bitnxt has covered the development in its report on Singapore's S$100 million tokenized MAS Bills.

Another relevant initiative involves Standard Chartered, OCBC and StraitsX.

The institutions tested interoperability between tokenized deposit networks, demonstrating how banks could coordinate payments across separate digital infrastructures.

Read Bitnxt's coverage of the Standard Chartered and OCBC tokenized deposit trial.

Although these projects address different financial processes, they share a common objective: reducing friction in institutional transactions through digital infrastructure.

JPMorgan's latest experiment adds another component to this trend.

Instead of focusing exclusively on how money moves, it examines whether software agents can responsibly decide when a transaction should be initiated.

This could become increasingly relevant as financial institutions adopt more advanced automation.

What JPMorgan and Hyundai Card Plan for 2027

The companies are targeting the first half of 2027 for a more advanced phase of testing.

The next stage is expected to involve multiple AI agents negotiating and coordinating transactions with one another.

This would be more complex than the initial treasury management experiment.

In the first phase, Hyundai Card's AI agent identified a funding requirement and submitted a transaction request.

Kinexys then evaluated the request against established authorization rules.

A more advanced model could involve agents representing different institutions, each operating under separate objectives and permissions.

For example, one agent might seek to purchase a financial service while another negotiates the terms under which it can be provided.

The agents would need to reach an agreement before an authorized payment could occur.

This introduces additional challenges involving negotiation rules, identity verification, liability and dispute resolution.

The companies have not announced a confirmed commercial launch date.

The 2027 timeline refers to further experimentation rather than the guaranteed availability of autonomous AI banking services.

Hyundai Card has also expressed interest in expanding the collaboration into other areas, including payments connected to vehicles.

Potential applications could include software-managed payments for mobility services.

However, the companies have not announced a production-ready in-vehicle payment service based on this experiment.

Could AI Agents Eventually Pay Other AI Agents?

The concept of AI agents independently conducting transactions is often described as agentic commerce.

In an agentic commerce system, software can act on behalf of an individual or organization to complete authorized commercial activities.

For example, an AI agent managing a company's cloud infrastructure might identify a need for additional computing resources.

It could compare available providers, evaluate pricing and request a purchase.

A second agent representing the service provider could negotiate the terms.

If both agents reach an agreement within their authorized limits, a payment system could process the transaction.

The important requirement is that the agents remain accountable to their respective owners.

An AI agent should not be able to create unlimited financial obligations simply because it has access to payment infrastructure.

This is where JPMorgan's work on signed mandates and institutional controls becomes relevant.

The technology could provide a framework for connecting autonomous decision-making with verified financial authority.

Nevertheless, several questions remain unresolved.

How should institutions handle an AI agent that makes an economically poor but technically authorized decision?

Who is responsible when an agent misinterprets a contract?

How can financial institutions prevent manipulated information from influencing automated transactions?

And how should disputes between independently operating agents be resolved?

These issues will become increasingly important as agentic commerce moves beyond controlled experiments.

What Are the Risks of Autonomous AI Financial Transactions?

The promise of AI-powered finance comes with substantial operational challenges.

Unauthorized transactions: Poorly designed permissions could allow agents to initiate transfers beyond their intended authority.

Incorrect financial decisions: An AI agent may select an unfavorable exchange rate or misinterpret account information.

Cybersecurity threats: Attackers could attempt to manipulate an agent's inputs, compromise its credentials or interfere with its transaction instructions.

Regulatory compliance: Banks must ensure that automated transactions satisfy applicable financial regulations, sanctions requirements and internal policies.

Accountability: Institutions need clear responsibility for actions performed by software acting on their behalf.

System reliability: AI services, banking infrastructure and communication networks must remain dependable during periods of heavy activity or disruption.

The JPMorgan experiment addressed several of these risks through identity verification, signed mandates, predefined limits and auditability.

However, the controlled test did not establish how the system would perform under all real-world conditions.

Further testing will be necessary before institutions can rely on such systems for large-scale financial operations.

Bitnxt Take: AI Agents Could Become the Next Layer of Blockchain Banking

At Bitnxt, we believe the JPMorgan and Hyundai Card experiment is important because it addresses a question that the financial industry will increasingly face: what happens when software begins initiating financial transactions rather than simply recommending them?

For years, banks have invested in digital payments, automated settlement and blockchain-based infrastructure.

Those developments have focused primarily on improving how transactions are processed.

The latest experiment explores a different opportunity.

It asks whether AI agents can manage parts of the decision-making process while remaining accountable to the institutions they represent.

In our view, that distinction matters.

A bank can automate a payment without using artificial intelligence.

But an AI agent that monitors financial conditions, compares available options and determines when an authorized transaction is necessary introduces a more flexible operating model.

The most interesting part of the JPMorgan experiment is not simply that the agents completed a simulated currency conversion.

It is that the transaction was connected to a verified identity, an approved mandate and enforceable controls.

We believe these safeguards will be essential if agentic commerce is to gain acceptance among regulated financial institutions.

The project also highlights an important limitation.

No real money moved during the experiment.

That means the companies have demonstrated a controlled transaction workflow, not yet a commercially proven autonomous payment network.

We would also caution against assuming that blockchain technology is automatically necessary for every AI-driven financial transaction.

Existing payment systems can support extensive automation.

Blockchain infrastructure must demonstrate additional value through settlement efficiency, programmability, auditability or interoperability.

The combination of AI and blockchain becomes more compelling when both technologies solve clearly defined problems.

We see a possible connection between this experiment and the wider growth of tokenized finance.

As more banks explore tokenized deposits, digital securities and wholesale CBDCs, AI agents could eventually become interfaces for managing transactions involving these instruments.

For example, an authorized treasury agent might monitor liquidity and arrange a tokenized deposit transfer or collateral movement when predefined conditions are satisfied.

That remains a future possibility rather than a capability demonstrated in the Hyundai Card trial.

Bitnxt's view is that the next stage of financial automation will depend less on giving AI agents unrestricted independence and more on creating trusted systems that allow them to act safely within clearly defined limits.

JPMorgan and Hyundai Card have taken an early step toward that model.

The more important test will come when multiple agents begin negotiating with one another and executing real transactions under institutional oversight.

What Happens Next?

The next major development is expected during the first half of 2027, when JPMorgan and Hyundai Card plan to explore more complex multi-agent transaction scenarios.

Future testing could provide greater insight into how independently operating AI agents negotiate, verify permissions and coordinate payments.

Another important milestone will be the publication of measurable performance results.

The companies have not disclosed transaction throughput, cost savings or commercial settlement performance from the initial experiment.

Those figures would help financial institutions evaluate whether the technology provides meaningful advantages over existing systems.

The development of regulatory and governance frameworks will also be important.

Banks must establish how autonomous financial decisions can be monitored, challenged and audited.

Finally, the industry will be watching whether Hyundai Card expands the technology into additional business applications, including corporate treasury operations and vehicle-related payments.

Conclusion

JPMorgan and Hyundai Card have successfully completed the first phase of a proof-of-concept involving autonomous AI agents and blockchain-based financial infrastructure.

Using Kinexys, the companies demonstrated how AI agents could monitor account balances, compare foreign exchange quotes and execute authorized simulated currency transactions without direct human intervention.

The experiment also incorporated identity verification, digitally signed mandates, transaction controls and auditability.

No real funds were transferred, and the technology has not yet been launched as a commercial banking service.

The companies are targeting further multi-agent testing in the first half of 2027.

For Bitnxt, the development highlights an emerging direction in digital finance, where artificial intelligence may increasingly interact with blockchain-based payment and settlement infrastructure.

The long-term opportunity is significant, but widespread adoption will depend on security, accountability, regulatory compliance and demonstrated commercial benefits.

Suggest a correction or update

Spotted something missing, outdated or incorrect on this page? Tell us what needs updating and share an official source or supporting link so our team can verify it.

More AI & Blockchain News

Share:

Bitnxt uses cookies to run the site, understand how it is used, and measure ads. You can accept or decline below, and change your choice any time in our Privacy Policy.