JPMorgan Chase has blocked outside artificial intelligence agents from accessing its customer accounts. Chief Executive Jamie Dimon announced the decision on the bank's third-quarter earnings call, where he identified third-party AI tools as a new class of operational risk for the largest United States bank by assets.

Dimon's objection centered on accountability. When customers authorize an outside AI tool to act on their account, he said, a transfer of liability occurs — moving responsibility for those actions from the customer and the bank to an unregulated intermediary. He described the danger in a single phrase: the risk of "rogue agents" — AI programs operating on a user's behalf without the bank's ability to verify or control what they do.

The statement carries weight because of where it was made. A third-quarter earnings call is a venue reserved for financial results and material strategic questions, and Dimon chose it to draw a line on third-party AI access. That placement signals the bank treats AI-agent risk as a board-level concern rather than a niche technology issue.

What does the policy actually change?

Under the block, JPMorgan customer accounts are no longer accessible to external AI agents — software programs that can take actions on a user's behalf, such as initiating payments or moving funds, rather than merely answering questions. The restriction applies to outside tools authorized by customers, not to AI systems the bank runs internally.

The decision turns on the liability question Dimon raised. Once a customer grants an outside agent permission to act, the bank can no longer distinguish between an instruction the customer intended and one a malfunctioning or compromised agent executed. Each such authorization, in Dimon's framing, shifts liability onto the bank for actions it did not control and could not audit.

For JPMorgan's roughly 60 million digital customers, the immediate effect is that third-party AI assistants cannot be plugged into their accounts to transact on their behalf. Customers who attempt to authorize such tools will find the access denied.

Why does the liability question matter?

An AI agent differs from earlier digital channels in one decisive respect: it acts autonomously. A customer using a banking app issues each instruction personally. A customer using an outside agent delegates decision-making to software built by a third company, operating outside the bank's security perimeter.

That structure creates the exposure Dimon flagged. If a rogue agent — one acting outside the customer's intent — moves money or executes trades, the question of who bears the loss has no settled answer. The bank did not process the instruction knowingly; the customer did not issue it. Dimon's answer is to close the channel until the liability question is resolved.

The move also reflects a defensive calculation. JPMorgan invests heavily in its own AI capabilities, and blocking outside agents protects both the customer relationship and the bank's data from intermediaries it cannot supervise. The decision converts an open question — who is responsible when an autonomous tool errs — into a firm policy: not on this bank's rails.

What comes next?

Dimon's public stance sets a marker for the industry. If the liability logic holds — that customer authorization of outside AI tools transfers risk to the bank — other large lenders face the same exposure and the same choice. JPMorgan, as the first major United States bank to state the position on a public earnings call, has framed the debate its competitors will now have to answer.

The third-quarter call thus marks a turning point in how banks treat AI agents: not as an extension of digital banking, but as counterparties whose authorization creates risk the bank did not agree to hold. The block is the enforcement of that position, stated in plain terms by the chief executive himself.