Citigroup and Bank of America Restrict AI Agents to Protect Deposits
Citigroup and Bank of America have restricted AI agents from accessing customer accounts to prevent massive deposit outflows caused by automated high-yield searching. Major banks are blocking third-party AI assistants from logging into customer accounts or directly executing transactions via digital platforms. This development highlights a major friction point between autonomous AI agents and traditional banking infrastructure, as automated tools threaten banks' low-cost funding models. If AI agents systematically move household savings from low-interest checking accounts to high-yield alternatives, banks could face severe profit squeezes and liquidity challenges. Bank of America CEO Brian Moynihan emphasized that outbound funds must require manual authorization, while JPMorgan Chase CEO Jamie Dimon cited the risk of thousands of uncontrolled agents. Meanwhile, banks still allow customers to use limited, read-only open banking services for account information.
## BACKGROUND
Traditional retail banks rely heavily on low-interest checking and savings accounts—known as low-cost deposits—to fund loans and generate profits. The rise of agentic AI, such as Meta's Muse personal assistant, enables automated systems to actively scan and move funds to accounts offering significantly higher interest rates, disrupting traditional banking dynamics.