Technology news and discussion on TalkFuse.
JPMorgan is working on AI agent transactions that don’t involve humans
JPMorgan has completed an initial trial of technology that lets AI agents carry out financial transactions without direct human involvement. The bank has not set a date for commercial launch, and says the harder challenge is establishing accountability and controls, which will be crucial if autonomous agents become more common in financial services.
The bank’s Kinexys unit limits agents to predefined transaction rules, such as blocking an agent authorised to spend up to $5 million from exceeding that amount. AvePoint reported that unauthorised autonomous actions were the third most common AI agent-related security breach, affecting 34.1% of companies that reported such incidents. Experts warn that agents could pass errors between one another, and say safeguards should include spending limits, restricted access, user monitoring and the ability to revoke permissions.
In short
- JPMorgan has completed an initial trial of autonomous AI payments.
- Commercial launch timing has not been set.
- Transaction limits and oversight are key safeguards.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Allowing AI agents to transact within tightly defined limits could make routine financial processes faster and more efficient, while freeing people to focus on decisions that need judgement. Pre-set spending caps, restricted access, monitoring and revocable permissions may contain risks, and trials could help banks develop controls before wider use.
The case against
Financial transactions made without direct human involvement raise questions about who is accountable when an agent makes a mistake or acts beyond its intended purpose. Errors could also spread between interacting agents, while reported incidents of unauthorised actions suggest safeguards may not always prevent harm; delaying deployment until oversight is demonstrably robust would protect customers and trust in the financial system.
Explained simply
AI agents are software systems that can carry out tasks with limited human input, including making payments under set instructions. In finance, that could let businesses automate routine transactions, but it also raises questions about who is responsible if an agent makes a mistake.
JPMorgan is a large US bank, and Kinexys is its unit focused on financial technology and digital payments. Banks would need controls such as spending limits and restricted access to keep automated transactions within agreed boundaries.
The issue matters because an error could spread if agents pass information or instructions to one another. Clear oversight and a way to stop an agent’s access are important considerations as financial firms explore using them.