US Banks Are Blocking AI Agents From Customer Accounts, Fearing a Deposit Drain

October 17 — The US banking industry has warned regulators that AI agents may automatically move idle funds into other accounts to earn users higher returns, draining a large amount of deposits. According to a report by the Financial Times on the 16th, Citigroup and Bank of America have also moved to restrict AI agents’ access to customer accounts.

Citi says that to protect customer accounts and personal information, the bank bans third-party AI agents or assistants from logging into customer accounts, and does not allow them to carry out transactions directly through Citi’s digital banking platform.

Bank of America chief executive Brian Moynihan revealed that his bank has also restricted AI agents’ access to individual customers’ deposits and is drafting rules to clarify where final responsibility lies for AI-authorised payments: “We firmly believe that money leaving an account must be authorised by a human. We are talking to the agent providers and telling them: ‘You surely don’t want a consumer to come back and complain that money was moved out of the account without knowing why, or where it went, only to find out an AI agent did it.’”

A person familiar with the matter confirmed that Bank of America has imposed restrictions.

In September this year, after Meta launched its Muse personal AI assistant, the banking industry’s worry about deposit flight intensified. Banks fear that AI agents will hunt for higher-yielding deposit accounts on consumers’ behalf and move large amounts of idle money out of existing accounts.

At the same time, AI companies are rushing out new tools and chatbots that help consumers manage their personal finances.

A report by Torsten Slok, chief economist at Apollo Global Management, added to the industry’s concerns. He warned that AI assistants could automatically move household money out of checking accounts, where the US average rate is just 0.1%, into other accounts paying 3.3% to 5%.

JPMorgan Chase, Bank of America and Wells Fargo together hold US$1.6 trillion (about 10.72 trillion yuan at current exchange rates) in non-interest-bearing deposits, 16% of the three banks’ total liabilities. If they paid depositors 3% interest, the three banks would have to spend an extra US$47 billion a year (about 314.8 billion yuan) — close to half their combined annual profit.

If banks had to raise individual deposit rates to stop customers moving money to higher-paying institutions, their profits could shrink sharply.

The report notes that if every household used AI agents to earn more on its cash deposits, banks could lose much of the low-cost deposits they use to make loans, which would create problems for the financial system as a whole.

A person familiar with the matter said Wall Street executives have also raised concerns with Federal Reserve supervisors about consumers increasingly using AI agents to handle financial transactions.

Regulatory officials have privately welcomed the banks’ moves to block AI agents from customer accounts. They believe the law may need to clarify how consumers should authorise AI agents to conduct financial transactions, and who is liable when a transaction goes wrong.

Citi says customers can still authorise some open-banking providers to obtain account information on a read-only basis within a limited scope.

JPMorgan Chase chief executive Jamie Dimon announced this week that, because the bank does not want thousands of potentially runaway agents, it will restrict AI agents’ access to customer accounts. “We must make sure we protect customers properly, so we are developing the tools to do that. In the meantime, we will not let any external AI agent roam freely inside the bank’s systems, because the risk to customers and to JPMorgan is too great.”

Christoph Stegemeyer, a senior partner at the consultancy Simon-Kucher, believes banks must take steps to encourage customers to keep their deposits and find ways to make deposit products stickier. Specific measures could include expanding loyalty programmes and using them to offer different lending rates.