Bank of America Tightens AI Agent Access, Warns Customers Not to Share Banking Passwords

Bank of America Tightens AI Agent Access, Warns Customers Not to Share Banking Passwords

Bank of America this week advised customers not to hand their banking login passwords to AI assistants. The bank has also limited third-party AI agents' account permissions to read-only, meaning the software can view account information but cannot move funds on the customer's behalf. According to an Axios report dated October 22, 2026, JPMorgan Chase, Morgan Stanley, Citigroup and Wells Fargo are taking similar action.

Approaches vary across institutions. Morgan Stanley has directly blocked third-party AI agents from accessing accounts; the other banks favor directing agents to regulated open banking channels rather than allowing them to use raw usernames and passwords. The banks note that sharing login credentials with third-party AI agents itself violates terms of service, and that account activity beyond the scope of user authorization has already been observed—including agents calling bank customer service directly.

JPMorgan CEO Jamie Dimon said on the company's third-quarter earnings call on October 13, 2026, that there may be "thousands" of out-of-control AI agents in the banking system. There have been no reports of widespread fraudulent charges or confirmed losses linked to these agents.

Liability is the focal point of the dispute. The banks have made clear that if a customer gives credentials to an AI agent and the agent carries out unauthorized actions, the customer bears the loss. This stance places risk on the user side and pushes agent-initiated payments toward authorization channels that can be monitored and held accountable.

Source: https://cryptobriefing.com/bank-of-america-ai-assistants-passwords/