Citigroup and Bank of America have restricted access for artificial intelligence (AI) agents to their clients’ accounts amid concerns that such tools could move large sums of idle cash in pursuit of higher returns, potentially destabilizing the banking sector. These measures come as AI personal assistants designed to help consumers manage finances rapidly enter the market.
Citi stated that it does not allow “third-party AI agents or assistants to log into a customer’s account or transact directly within our digital banking platforms.” The bank said this policy is intended to protect customer accounts and information. Meanwhile, Bank of America’s CEO Brian Moynihan indicated that the lender is limiting AI agents’ access while it develops protocols to clarify liability for payments authorized by AI systems. “We believe there should be a human in the loop to authorise money moving out of your account,” Moynihan said, underlining the bank’s caution about automated transactions. A source close to the matter confirmed the bank’s decision to restrict agent access.
The concern follows the launch of Meta’s Muse AI personal assistant in September, which intensified worries that AI bots could automatically transfer consumers’ idle cash from low-interest checking accounts to higher-yielding alternatives. This could lead to significant outflows from banks’ non-interest-bearing deposits, a primary source of low-cost funding.
A report by Apollo’s chief economist, Torsten Sløk, highlighted the potential impact: AI agents could sweep household balances into accounts offering between 3.3 and 5 percent interest, compared to the roughly 0.1 percent average on checking accounts. Major U.S. banks including JPMorgan Chase, Bank of America, and Wells Fargo currently hold $1.6 trillion in such deposits, representing about 16 percent of their liabilities. Meeting a 3 percent interest cost on this amount would total approximately $47 billion annually, nearly half of their combined yearly earnings.
Sløk noted that widespread use of AI to optimize cash returns could induce banks to lose a substantial portion of these inexpensive deposits, which underpin their lending operations and financial stability. Executives from these institutions have reportedly discussed these concerns with Federal Reserve regulators, although the Fed has declined to comment publicly.
Regulators are said to privately support the moves by banks to block AI agents’ direct access to accounts. JPMorgan CEO Jamie Dimon also announced that the bank would restrict AI agent access, citing the risk of “rogue agents” operating at scale.
As AI tools continue to develop, financial institutions are balancing the potential benefits of innovation against risks related to security, liability, and systemic stability. The evolving stance among banks suggests a cautious approach toward integrating AI-driven automation in retail banking transactions.