Citigroup and Bank of America are the latest U.S. financial institutions to restrict third-party AI agents from accessing customer accounts. Citi has made clear that it does not allow third-party AI agents or assistants to log in to customer accounts or conduct transactions directly on its digital banking platform, in order to protect the security of customers’ accounts and information. Bank of America CEO Brian Moynihan also said the bank is restricting AI agents’ access to consumer deposits, emphasizing that a person must be directly involved and authorize any transfer before funds can leave an account, to prevent uncontrolled, unauthorized transfers.
Banks and market analysts are chiefly concerned about deposit outflows and reduced profits. With various personal AI agents being launched, there are concerns that they could automatically move idle cash earning low interest into higher-yielding accounts. Torsten Slok, chief economist at Apollo, notes that JPMorgan Chase, Bank of America, and Wells Fargo currently hold about $1.6 trillion in non-interest-bearing deposits, equal to 16% of their liabilities. If AI-driven cash optimization forces the banks to raise deposit rates to 3%, the three banks would incur roughly $47 billion in additional annual interest costs, directly undermining the low-cost funding base they rely on for lending.
Regulators have privately welcomed banks’ restrictions on AI agents, while noting that legislation will eventually be needed to clarify the customer authorization process and who is liable when transactions go wrong. JPMorgan Chase CEO Jamie Dimon has also said that external AI agents will not be allowed unrestricted access to the bank’s systems until it has developed secure tools of its own. In response to the potential liquidity shocks posed by the technology, industry consultants say banks will rely increasingly on incentives such as customer loyalty programs to encourage deposit stickiness and stabilize their funding base.