Major U.S. banks are limiting the activity of AI agents that access customer accounts. Concerns are growing over the safety of financial transactions and where responsibility lies after signs emerged that some AI agents tried to act beyond the scope of user instructions.

TechRadar reported on Oct. 27 that major U.S. financial institutions including JPMorgan Chase, Bank of America and Morgan Stanley have begun controlling third-party AI agents’ access to customer accounts.

The issue differs from how hackers break into bank systems. It arose after some customers directly provided account login details to AI agents to check financial information or manage assets. The explanation is that even if users allow access, that does not mean banks have approved the AI tool’s connection or transactions.

Some banks have reported suspected cases in which AI agents acted autonomously, such as contacting customer service without an explicit request from the user. No cases have been confirmed in which AI transferred funds on its own or caused actual financial damage.

Banks are blocking third-party AI agents or limiting what they are allowed to do. Morgan Stanley blocked external agents from accessing its platform, but said it did not find any cases of an AI agent acting without approval accessing its systems.

JPMorgan Chase CEO Jamie Dimon (제이미 다이먼) said it could pose risks to customers and the bank, and said the bank will not allow access by unverified third-party AI agents.

A key concern for the financial industry is AI’s autonomous decision-making. Finding unnecessary subscription services when a user asks to cut spending is a completely different issue from actually cancelling payments or moving funds.

The outlet argued that permissions to view financial information should be separated from permissions to move money, and that users should provide explicit approval each time for actual transactions such as remittances.

Who would be responsible if an AI agent makes a wrong decision and causes losses also remains a task. The banking industry is considering allowing AI to read financial information while restricting authority to execute transactions directly.