US Banks Restrict AI Agents Over Deposit Flight and Liability Fears
Major US banks, including Citigroup and Bank of America, are restricting AI agents from customer accounts to prevent billions in deposit outflows to higher-yielding alternatives.
Major US banks are beginning to erect barriers against artificial intelligence agents attempting to operate within customer accounts, fearing that the technology could lead to the outflow of billions of dollars from bank deposits into savings vehicles offering higher yields. According to a report in the Financial Times, Citigroup and Bank of America have joined lenders restricting AI agents' activity in customer accounts. The move reflects growing concern in the banking system over autonomous tools capable of managing client funds, comparing returns, and transferring money between accounts without requiring human intervention for every action.
Citigroup clarified that it will not allow artificial intelligence agents or external digital assistants to connect to customer accounts or execute direct actions on its digital banking platforms. The bank explained that the restrictions are designed to protect customer accounts and personal data. However, bank clients will still be able to authorize certain open banking service providers limited access to view information in their accounts, without the ability to execute transactions.
Bank of America is also restricting AI agents' access to customer deposits, at least until regulations are established determining who bears liability in the event that an artificial intelligence system executes an erroneous or unauthorized funds transfer. Bank CEO Brian Moynihan stated on Wednesday to CNBC that he believes a human must be involved in approving money transfers from a bank account. "We firmly believe there needs to be a human in the process to approve taking money out of the account," Moynihan said. According to him, the bank is working with AI agent developers to prevent a situation where customers discover their money was transferred without understanding why or where.
Banking system concerns intensified following the launch of Meta's Muse artificial intelligence personal assistant in September. Technology companies are developing tools at an accelerated pace designed to help consumers manage their money, and banks worry that such agents could identify within seconds accounts where money yields virtually no interest and transfer it to more lucrative avenues. The financial risk to banks could be significant. Torsten Sløk, chief economist at investment giant Apollo, recently warned that AI agents could automatically transfer household funds from checking accounts, where the average US interest rate stands at about 0.1%, to accounts offering 3.3% to 5% interest.
According to data presented by Sløk, three of the largest US banks – JPMorgan Chase, Bank of America, and Wells Fargo – together hold $1.6 trillion in non-interest-bearing deposits, representing about 16% of their liabilities. If forced to pay 3% interest on these deposits to prevent funds from moving to competitors, their annual cost would reach about $47 billion – an amount equal to nearly half their aggregate annual profit. "If every household uses AI agents to maximize the yield on its cash balances, banks could lose a large part of the cheap deposits they rely on for lending, and this could be a problem for the entire financial system," Sløk wrote.
According to the report, Wall Street executives have already approached regulators at the Federal Reserve expressing concerns over the rapid rise in the use of AI agents to execute financial operations on behalf of customers. The Fed declined to comment, but regulatory officials expressed support in closed conversations for the banks' steps to block access to accounts. In their view, legal clarification is required regarding how clients grant authorization to autonomous systems to perform actions on their behalf, and who bears responsibility when something goes wrong.
JPMorgan Chase CEO Jamie Dimon also announced this week that the bank will restrict external AI agents' access to customer accounts. "We don't want agents acting without control, and there could be thousands of them," he said on an earnings call following the bank's third-quarter results. According to Dimon, the bank is developing tools to protect its clients, but until those are available, he does not intend to allow any external agent into the bank's systems. "It's simply too dangerous for customers and for JPMorgan Chase," he clarified.
Alongside access restrictions, banks will also have to contend with the business threat posed by the technology. Christoph Stegmeyer, a senior partner at consulting firm Simon-Kucher, estimates that banks will be forced to find new ways to incentivize customers to keep their savings with them, including expanding loyalty programs and interest benefits on loans. According to him, to cope with an era where artificial intelligence can easily transfer funds between financial institutions, banks will need to make their deposit products more attractive and make departure more difficult.
