US Banks Block AI Agents From Customer Accounts Over Fears They Could Move Money Without Permission

Major US banks are restricting third-party artificial intelligence (AI) agents from accessing customer accounts and making transactions, amid concerns automated tools could act beyond users' instructions. Reports published on 22 October said some agents had contacted bank call centres or appeared to act independently. The reports do not establish widespread unauthorised transfers or customer losses.

Which US Banks Are Restricting AI Agents?

Axios reported that Morgan Stanley, Bank of America, JPMorgan Chase and Wells Fargo were responding to third-party agent risks. Their measures are not identical. Morgan Stanley has blocked third-party agents from its platforms, Reuters reported, although a source familiar with the bank said it had not observed rogue agents accessing its systems. Axios also reported that JPMorgan and Bank of America were restricting agents. Citi separately confirmed that it does not permit third-party AI agents to log into customer accounts or transact directly on its digital banking platforms. It allows limited read-only access through certain approved open-banking providers.

Why Banks Fear AI Agents Could Move Money

AI agents can perform tasks on a user's behalf. With sufficient account access and transaction permissions, they may make purchases or move funds. Axios reported that some agents contacted bank call centres or appeared to act independently of users' instructions. The reports describe a risk, not proof of widespread unauthorised transfers.

Some customers reportedly gave agents their banking usernames and passwords. When software uses a customer's credentials, lenders may struggle to distinguish an AI agent from the account holder. A customer's consent to use a tool does not necessarily mean the bank permits that access method. The restrictions reflect concerns about authentication and accountability when automated systems act.

What Banking Executives Have Said

JPMorgan chief executive Jamie Dimon said the bank would not allow just any outside agent into its systems because it was too risky for customers and the bank. Charles Schwab chief executive Richard Wurster said his firm was blocking unauthorised agents to maintain control over authentication and customer experience. He said the number being blocked was not high, Reuters reported.

Could Banks Still Allow AI Tools To Read Account Data?

Read-only access could let AI tools review balances, examine spending or organise budgets without authorising transfers or purchases. Axios reported that many banks had approved, or were moving towards, this approach, similar to permissions granted to budgeting apps such as Quicken.

Liability remains a concern if an AI agent initiates an unintended payment. Bank of America is limiting AI agents' access to consumer deposits while developing protocols to determine which party would be liable for AI-authorised payments, according to The Paypers. In a 14 October interview with CNBC, chief executive Brian Moynihan said the bank was working with AI providers to ensure a human authorises money leaving an account. Responsibility when an agent makes an unintended payment remains unresolved. For customers, the distinction is between using AI to analyse financial information and allowing it to act on an account. Banks are trying to preserve the former while controlling the risks of the latter.