U.S. banks block unauthorized AI agents from moving customer funds

The crackdown on automated agents

Major U.S. financial institutions are taking a firm stance against the rise of third-party artificial intelligence agents. According to reports from pymnts.com, several large banks have begun detecting and deleting unauthorized AI agents that attempt to operate inside customer accounts.

These agents are software programs designed to perform tasks on behalf of a user. While they are becoming more common in various industries, banks are concerned about their ability to interact with sensitive financial data. Some detected agents have reportedly tried to execute actions that the customers never actually requested.

The pushback comes as the technology matures. Banks are prioritizing security and regulatory compliance, ensuring that automated tools do not create new avenues for fraud or unauthorized movement of capital.

Who is enforcing the rules

The crackdown involves some of the largest names in American finance. JPMorganChase, Citigroup, Bank of America, and Charles Schwab have all issued statements clarifying that AI agents are not permitted to move customer money. These firms are establishing clear boundaries between what an AI can do and what it cannot do.

Charles Schwab CEO Rick Wurster noted that the brokerage is blocking unauthorized AI agents a double-digit number of times every week. This suggests that the attempts to bypass security protocols are frequent enough to require active monitoring and intervention by the bank's technical teams.

Bank of America CEO Brian Moynihan emphasized a specific requirement for the future: a human must stay in the loop to approve any money moving out of an account. This 'human-in-the-loop' approach ensures that while an AI might suggest a move, a person must provide the final authorization.

Read-only access vs. transactions

There is a distinction between an AI agent simply looking at your data and an AI agent acting on your behalf. Currently, the four major firms mentioned do not allow third-party agents to log into accounts to perform transactions. However, they do allow these agents to perform read-only functions.

Read-only functions mean the AI can see information, such as your current account balance, but it cannot change anything or send money to another person. This allows users to still benefit from AI-driven insights—like getting a summary of their spending—without the risk of the AI making a mistake or being hijacked by a malicious actor.

Citigroup has specifically stated that it does not let third-party agents log into accounts to transact. This creates a safety buffer where the AI can be a helpful assistant, but never a primary spender.

The hurdles for automated finance

The goal for many developers is to create AI agents that can automatically move household cash to accounts with higher interest rates. This would remove the 'friction' of manual transfers, which often causes people to leave money in low-interest accounts. However, the logistics of these automated moves are currently very complicated.

U.S. Bancorp CEO Gunjan Kedia stated during an earnings call that the bank has not encountered these types of third-party cash-sorting agents yet. He noted that because the logistics are so complex, the bank does not expect this to be a near-term impact on their operations.

PNC Financial Services Group Chairman and CEO William Demchak highlighted the legal and safety barriers remaining. He argued that for agents to eventually move money, there would need to be new laws passed regarding liability shifting. He also noted that such a system would require highly secure connections and authentication to be viable.

What this means for users

For the average consumer, this means that the dream of a fully autonomous 'money manager' AI is not going to happen overnight. While you may soon see more tools that can analyze your spending habits or suggest where to put your savings, you will still be the one clicking the 'send' button.

Banks are positioning themselves as the gatekeepers of security. By blocking these agents, they are attempting to prevent a scenario where a software glitch or a hacked AI could drain a customer's account in seconds. For now, the industry is moving toward a hybrid model where AI provides the intelligence, but humans provide the permission.