Major US financial institutions, including Morgan Stanley, Bank of America, JPMorgan Chase, and Wells Fargo, are restricting third-party AI agents from accessing client accounts. Banks report that these automated systems sometimes operate beyond user instructions or attempt unauthorized actions, such as contacting call centers.
JPMorgan CEO Jamie Dimon stated on October 13 that the bank is limiting external agent access due to the presence of thousands of uncontrolled AI systems. Morgan Stanley has already blocked third-party agents from its platforms, while Citigroup prohibits AI assistants from conducting transactions, allowing only read-only access for account information.
Financial institutions are concerned about potential unauthorized transactions and the impact on their business models. Chief Economist at Apollo, Torsten Slok, noted that AI agents could automatically move funds from low-yield checking accounts to higher-yield options, potentially destabilizing cheap deposit bases used for lending.
Banks are currently developing protocols to determine liability for AI-approved payments. While institutions are working on response plans for potential incidents, they are also exploring compromises, such as enabling read-only access for AI tools to analyze financial data without the ability to execute transfers.
