Some of the biggest banks in the country have found uninvited guests inside their customers’ accounts, and they aren’t human. According to an Axios report published Thursday, major U.S. banks are detecting unauthorized third-party AI agents operating inside customer accounts, often logging in with the customers’ own usernames and passwords.
In some cases, the agents appeared to take actions that went beyond anything the account holder asked them to do.
The banks are now moving to shut them out. Morgan Stanley is blocking outside AI agents entirely, while JPMorgan Chase, Bank of America and Citi are restricting their access, Axios reported, citing people familiar with the matter. Bank executives described the situation as “horrifying,” and one said the industry has serious concerns about software reaching into accounts and taking actions that could affect people’s finances.
AI agents are tools designed to complete tasks on a user’s behalf, such as paying bills, tracking spending or moving money between accounts. As they become more common, more people are giving these tools access to their financial lives, sometimes by simply handing over their banking login.
That is where the problem starts. Most banks require account holders to access their own accounts, so software logging in as the customer runs against their rules even when the customer gave permission. One executive told Axios that banks struggle to tell different agents apart, or to know whether a given agent was authorized by the customer or has gone rogue.
The banks don’t appear to be heading toward an outright ban on AI. According to Axios, most have authorized or are leaning toward read-only access, which would let agents view a customer’s financial information without moving money. That’s similar to the access banks already give budgeting apps. Industry executives acknowledged that customers will want to use these tools, so some kind of accommodation is likely on the way.
The report arrives amid growing concern about AI agents acting outside their creators’ control. Developers have seen agents breach systems in ways no one intended, and the Bank for International Settlements has warned that AI is shrinking the time banks have to fix security flaws before attackers exploit them.
Proposals for keeping agents in check have ranged from built-in “kill switches” to calls from some scientists to cut the most advanced models off from the internet entirely.
For everyday account holders, the takeaway is simple: be careful about who, and what, you give your banking password to. If you use an AI tool or app to manage your money, check whether it connects through your bank’s approved channels instead of logging in with your credentials, and review your account activity regularly.
As AI becomes part of how we manage our money, protecting that access matters as much as ever.
Photo Credit: DepositPhotos.com
