Big U.S. banks have begun restricting AI agents' access to customer accounts
The largest American banks are strengthening protection of customer accounts from third-party artificial intelligence agents. Financial institutions detect automated programs that log into banking systems using user credentials, and sometimes attempt actions that may exceed the received instructions.
Axios reported this on October 22, citing sources in the banking industry. The problem concerns, in particular, Morgan Stanley, JPMorgan Chase, Bank of America, Wells Fargo, and Citigroup.
According to the publication, Morgan Stanley has already begun blocking third-party AI agents, and JPMorgan, Bank of America, and Citi are imposing restrictions on their access. Banks are trying to prevent situations where automated systems gain the ability to independently carry out operations with customer funds.
AI agents differ from regular chatbots in that they can not only provide answers but also perform sequential actions on the internet on behalf of the user. For example, open websites, fill out forms, make purchases, or interact with financial services.
In some cases, bank account holders voluntarily passed logins and passwords to such programs. However, a client's permission does not always mean the access complies with bank rules. Most financial institutions require the user to log in to the personal account independently or use specially agreed data exchange mechanisms.
Separate concerns are raised by cases where AI agents, according to Axios sources, called bank contact centers or attempted actions without an obvious instruction from the account holder. For banks, the difficulty lies in not always being able to distinguish a human from an automated program and determine the scope of authorities granted to it.
However, the reports do not contain confirmed cases of theft of funds as a result of such actions. A Morgan Stanley representative told Axios that the bank had not detected third-party AI agents acting contrary to established rules on its platforms. Thus, the imposed restrictions are primarily preventive in nature.
Some banks began restricting third-party AI systems even before the new Axios report. In particular, on October 13, JPMorgan Chase CEO Jamie Dimon said during the quarterly results presentation that the bank would not allow third-party agents to work with customer accounts without control.
As reported by the specialized publication Finextra, JPMorgan considers the use of banking logins and passwords by third-party programs dangerous. At the same time, the company supports secure data exchange through authorized financial service providers.
One of the possible compromises is allowing AI agents access only to view information. Such a mechanism allows automated assistants to check balances or analyze expenses, but does not give them the right to transfer money or pay bills.
The brokerage company Charles Schwab adheres to a similar policy. Its CEO Rick Wurster told The Wall Street Journal about a cautious approach to using autonomous systems in financial management. The company also restricts unauthorized access by third-party agents.
For banks, the development of such technologies creates not only cybersecurity risks but also new questions of liability for financial operations. If a client themselves gives access to an AI system and it performs an unwanted action, determining the responsible party may become complicated.
According to Axios, American banks are already developing response scenarios for possible large-scale incidents, including mechanisms for rapid blocking of dangerous activity and restoring customer access to accounts.
The financial industry recognizes that demand for automated banking services will grow. Therefore, institutions face the task of ensuring AI use without transferring unlimited control over customer funds to third-party programs.
For more information: Axios, Finextra, The Wall Street Journal
