Skip to content
Cisent Message us on WhatsAppMessage us

Blog & news

Threat researchCisent

Big US banks are finding AI agents inside their customers' accounts

Customers hand their username and password to AI assistants, and banks can't always tell whether what the agent does was the customer's request.

On October 22, 2026, Axios reported that big US banks are finding third-party AI agents operating inside their customers’ accounts.1

What happened

  • The agents log in with the customer’s own username and password, sometimes with the customer’s permission. Some attempted actions that didn’t seem to come from a customer’s instruction, and others called the bank’s call centers.1
  • Morgan Stanley has begun blocking third-party agents, and JPMorgan, Bank of America and Citi are restricting them. Most banks are moving toward read-only access, which lets agents see the data without acting on it.1
  • According to Axios’s sources, many banks can’t tell whether an agent’s activity was directed by the customer.1
  • Chase reminded customers not to give their username and password to anyone. Morgan Stanley said it hasn’t seen rogue AI agents on its platforms.1

Why it matters

A valid login shows that someone got in, not who decided each action. And the habit that creates the risk, handing your password to a tool so it can run errands, isn’t in the system: it’s in people.

Where Cisent comes in

Not handing your password to an agent is learned: it’s a habit, not a patch.

Cisent Iroh is the agent that trains your team. It talks with people, explains attack techniques and response measures in plain language, and teaches how to respond, for technical and non-technical staff alike. It’s an add-on to any Sentinel product.

Explore Cisent Iroh

Sources

  1. Madison Mills, Axios, “Scoop: Unvetted AI agents are accessing U.S. bank accounts”, October 22, 2026. https://www.axios.com/2026/10/22/ai-agents-banks-unauthorized-accounts-access↩︎