Citigroup and Bank of America have joined JPMorgan Chase in limiting how autonomous AI tools interact with consumer bank accounts, according to the Financial Times. The restrictions come as US lenders tell regulators that AI agents could move idle deposits into higher-yielding products on customers’ behalf. Questions over liability for AI-authorised payments also remain unresolved.
Citi has stated that it does not allow third-party AI agents or assistants to log into customer accounts or carry out transactions directly within its digital banking platforms. The bank said the policy is designed to safeguard customer accounts and information. Customers can still grant limited, read-only access to their account information to Certain Open Banking providers, so data sharing remains possible while transactional access is closed.
Liability and human authorisation
In an interview with CNBC on 14 October 2026, Bank of America indicated that the bank is limiting AI agents’ access to consumer deposits. The restriction will stay in place while the bank develops protocols to establish which party is ultimately liable for payments authorised by AI. It was also mentioned that the bank believes a human should remain in the loop to approve money leaving an account, and it is working with agent providers to avoid cases where consumers cannot trace funds an agent has moved.
During the bank’s third-quarter 2026 results call, JPMorgan Chase said the bank would also block AI agents from accessing client accounts, citing the risk of ‘rogue agents’ that could number in the thousands. JPMorgan is building its own tools to protect customers and, until they are ready, will not allow outside agents to operate within its systems.
Deposit outflow concerns
The moves follow Meta’s launch of its Muse AI personal assistant in September 2026. The launch added to concerns that consumer-facing bots could shift large volumes of idle cash out of bank deposits in pursuit of better returns. AI companies are also introducing a growing number of tools aimed at personal finance management.
A report by Apollo’s chief economist warned that AI assistants could automatically sweep household cash into accounts paying between 3.3% and 5%. The national average rate on US checking accounts is 0.1%. JPMorgan, Bank of America, and Wells Fargo together hold USD 1.6 trillion in non-interest-bearing deposits, equal to 16% of their liabilities. Paying a 3% rate on those balances would cost the three banks USD 47 billion a year, close to half of their combined annual earnings. The economist argued that widespread use of such agents could erode the low-cost deposit base banks use to fund lending, which would affect the wider financial system.
Regulatory dimension
According to a person familiar with the discussions, Wall Street executives have contacted the Federal Reserve about the rapid rise in consumers using AI agents for financial transactions. The Fed declined to comment. Regulators have privately welcomed the banks’ restrictions. Furthermore, officials believe legal clarification may be needed on how customers authorise autonomous systems to transact on their behalf, and on who is liable when something goes wrong.
The restrictions put access control at the centre of the debate over agentic finance in the US. Banks are separating read-only data access from transactional permissions while liability frameworks remain undefined. A senior partner at consulting firm Simon-Kucher said banks will need to give customers reasons to keep savings with them, for example, expanded loyalty programmes linked to the borrowing rates on offer. In practice, that means making deposit products ‘stickier’.
