Financial Services
Citi and Bank of America restrict artificial intelligence agents to protect customers, and deposits
Source: Financial Times, 16 Oct 2026
Citigroup and Bank of America have restricted artificial intelligence agents from accessing client accounts, the Financial Times reports today. Citi said it does not permit third-party agents or assistants 'to log into a customer's account or transact directly within our digital banking platforms'. The banks have told regulators they fear agents will move idle cash to higher-yielding accounts: JPMorgan Chase, Bank of America and Wells Fargo hold $1.6tn of non-interest-bearing deposits, and paying 3 per cent on them would cost $47bn a year, almost half their collective annual earnings.
Customer protection is the stated reason; deposits are the arithmetic. Both can be true, and customers will see the gap. Brian Moynihan's argument, that a human should authorise money leaving an account until liability for an agent's mistakes is settled, is the one that stands up on its own. Blocking alone does not answer Torsten Slok's warning at Apollo: agents could sweep cash into accounts paying 3.3 to 5 per cent, against 0.1 per cent on the average checking account.
If you are a bank, state the protection case in terms a customer can check: who is liable if an agent errs, and what you permit instead, as Citi does with read-only access through open banking providers. Then give customers a reason to stay that does not depend on the gate; Christoph Stegmeier of Simon-Kucher says deposit products must become stickier. If you sell to banks, expect agent-access terms in your next security review.