AI assistants are starting to do more than answer questions. They can shop, compare prices and increasingly act on your behalf. Money is where things get complicated.
On October 17, the Financial Times reported that Citigroup had joined other major U.S. banks in restricting third-party AI agents from accessing customer accounts for transactions. Bank of America and JPMorgan Chase have also been cautious about letting autonomous agents move money.
Why banks are nervous
The basic concern is simple: an AI agent with financial access is not just software that recommends what to do. It may eventually be able to do it.
- Move idle cash to a higher-yield account.
- Pay a bill automatically.
- Choose between competing financial products.
- Potentially initiate transactions without a person clicking the final button.
That creates a difficult question: who is responsible if the agent gets it wrong?
This is bigger than a login problem
Banks already know how to authenticate customers. The harder problem is determining what authority a customer can safely delegate to an AI agent — and how a bank can distinguish a legitimate instruction from a compromised or rogue agent.
Regulators are paying attention too. Reuters reported on October 13 that U.S. and European banking watchdogs are increasing scrutiny of advanced and agentic AI, with particular attention to controls, cyber risk and the ability to recover when something goes wrong.
The tension nobody should ignore
There is also an economic incentive hiding underneath the security debate.
An AI agent designed to optimize your finances could notice that cash is earning almost nothing and move it somewhere paying more. That could be great for the customer — but costly for a bank that benefits from low-cost deposits.
So the coming debate will not be only about whether AI agents are safe. It will also be about who controls the financial relationship: the bank, the customer, or the software acting for the customer.
What this means for you
For now, be cautious about giving any third-party AI tool credentials or broad authority over a financial account. A useful AI assistant can analyze information and suggest an action without needing permission to move your money.
The useful line to remember: let AI help you decide before you let AI decide for you.
Agentic finance is likely coming. The important part is building the permission, liability and security rules before autonomous money movement becomes ordinary.
Sources: Financial Times, “Citi becomes latest US bank to restrict AI agents from accessing accounts,” Oct. 17, 2026; Reuters, “Banking watchdogs ramp up AI scrutiny as stability fears mount,” Oct. 13, 2026.
Leave a Reply