US banks are blocking AI agents. What would let one back in?
On 22 October 2026 Axios reported, citing people familiar with the matter, that Morgan Stanley, JPMorgan Chase, Bank of America and Citi are restricting third-party AI agents from customer accounts, for fear the agents could make transactions the customer did not intend.
As reported, banks have found agents signing in with customers' credentials, sometimes with the customer's permission but against the bank's terms, and some agents contacting call centres or acting on their own. The difficulty is telling an agent following the customer's instructions from one that is not. Morgan Stanley said it had blocked third-party agents despite not having seen rogue AI agents.
A bank that cannot answer "did the customer really ask for this?" will say no to every agent. So the question for anyone building agents that touch money: when your agent pays, what could you show the bank, and the customer, about what it was asked to do and what it actually did?
This is a reading of public reporting, not legal advice.
Source: PaymentExpert, 23 October 2026, reporting Axios. https://paymentexpert.com/2026/10/23/ai-agents-blocked-by-banks/