1 hour ago
9
JPMorgan Chase has a message for the AI agents of the world: you’re not on the list.
During its third-quarter earnings call on October 13, 2026, the bank said it is blocking outside AI agents from directly accessing customer accounts. Executives cited security risks, client protection and a growing worry about software acting on people’s behalf without enough guardrails.
What Dimon and Barnum actually said
CEO Jamie Dimon framed the decision as a matter of risk, for customers and for the bank itself.
“We’re not going to allow just any outside agent to come in inside the company because that’s just too risky for clients and for JPMorgan.”
He also warned about rogue agents. Think of these as AI tools that go off script, get hijacked, or simply do something the customer never intended.
Dimon raised a second concern: liability. When a customer hands an AI agent access to their account, responsibility for what happens next gets murky. If the agent moves money it shouldn’t, who eats the loss? The customer, the AI company or the bank?
CFO Jeremy Barnum added that the bank is working to prevent unauthorized access to customer information. He confirmed it is blocking third-party access “to the extent that we’re able to block those.”
Screen scraping, explained
Part of the concern centers on credential sharing, including a practice called screen scraping.
Picture giving a stranger your house keys so they can check whether your mail arrived. Screen scraping works similarly. A customer hands over their bank username and password, and a third-party tool logs in as them to read and pull account data.
JPMorgan has historically resisted this kind of access. The arrival of AI agents makes the old problem bigger. An agent doesn’t just read your balance. It can potentially take actions, which raises the stakes if something goes wrong.
Building in-house instead
JPMorgan isn’t swearing off AI. It’s just choosing to make its own.
The bank is developing internal agentic AI tools while sticking to strict security measures. “Agentic” here means AI that can carry out multi-step tasks on its own, not just answer questions in a chat window.
This isn’t the first time the bank has tapped the brakes on outside AI. In June 2026, JPMorgan restricted its Hong Kong staff’s access to Anthropic’s Claude, citing licensing and compliance issues.
Why this matters beyond JPMorgan
Industry analysts have cautioned against potential disintermediation, wherein consumer AI agents autonomously move capital without the bank’s oversight in pursuit of better interest rates or investment options.
There’s also a competitive angle hiding in plain sight. By blocking outside agents while building its own, JPMorgan keeps the customer relationship firmly in-house. Security is the stated reason, and it’s a legitimate one. Control over the customer experience is a convenient side effect.
What investors and customers should watch
A few open questions remain worth tracking:
How effective is the block? Barnum’s own wording suggests JPMorgan can’t stop every outside agent.
Who ends up liable? Dimon’s point about liability transfer is likely to come up again as regulators and courts catch up to agentic AI. Until the rules are clearer, banks have strong reasons to keep outside agents at arm’s length.
How good are JPMorgan’s own tools? Blocking competitors only works long term if the in-house product is worth using. The bank is betting it can deliver AI convenience without handing anyone else the keys.
For now, the position is clear. AI agents are welcome at JPMorgan, as long as JPMorgan built them.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.
English (US) ·