US banks block AI agents from operating in customer accounts

Major US banks are reportedly blocking third-party AI agents from operating inside customer accounts, with Axios noting that many banks allow only read-only access or are moving in that direction. The practice involves external AI tools that have been granted customers’ usernames and passwords to act on their behalf, a setup that banks say raises significant security and risk concerns.

Bank security implications are underlined by Bank of America’s public caution that customers remain responsible for activity conducted with shared credentials, highlighting the security and compliance risks of feeding such access into the banking system.

The discussion includes diverse perspectives. Bianco Research president Jim Bianco suggested that banks’ behaviour could be driven by revenue protection, arguing that AI agents might exploit deposits and trigger loss of income for traditional banking models.

JPMorgan Chase chief executive Jamie Dimon meanwhile emphasised that the bank would not allow arbitrary external agents into its systems because it is “too risky for clients and for JPMorgan.” Axios reported that many banks are already permitting, or moving toward, read-only access to allow analysis without the ability to transfer funds.

The article also notes that giving agents access to sensitive data—such as bank statements—remains a security concern, since compromised or manipulated agents could still disclose information or facilitate misuse. Consumers are urged to use approved connection methods, limit the data shared, and carefully review any agent-driven actions.