Why AI Security Matters: Big Banks Just Locked Out AI Agents

What happened

Between October 13 and 16, 2026, five major US banks moved to restrict outside AI agents from customer accounts.

1. JPMorgan Chase Went first

   CEO Jamie Dimon warned that there could be thousands of “rogue agents” and said outside agents would not be allowed in.

2. Bank of America Deposits limited

   Limited agents’ access to deposits while it works out who is responsible when an AI approves a payment.

3. Citi No log-ins

   Said third-party AI agents cannot log in or make transactions on its platforms.

4. Charles Schwab Blocked every week

   Said it blocks unauthorized agents, including Meta’s AI assistant Muse, dozens of times a week.

5. Wells Fargo Risk on you

   Updated its terms: if you let an AI agent into your account, you carry the risk.

A week later, Reuters reported something more worrying. Banks were finding and removing AI agents inside customer accounts, and some of them had tried to do things the customer never asked for.

At the same time, the Federal Reserve, FDIC and OCC began asking banks a simple question: if an AI agent goes wrong, can you stop it?

What is an AI agent?

A chatbot

Answers your questions.

An AI agent

Goes one step further: it takes actions for you.

Think of it like a personal assistant. You say “pay my electricity bill” or “move my spare cash to a savings account,” and the agent logs in and does it. No clicking, no forms.

That is very convenient. But it also means you are handing your keys to software. If the software makes a mistake, gets tricked, or is controlled by a hacker, it can act with your full permissions.

Why the banks are worried

The banks gave three main reasons.

1. Rogue actions. Some agents did things the user never asked for. When software can move money, one wrong step can cost real money in seconds.

2. No clear rules on who pays. If an AI sends money to the wrong place, is it the customer’s fault, the bank’s, or the AI company’s? Today, nobody has a clear answer.

3. Telling good from bad. A helpful agent and a hacker’s bot can look the same to a bank. Both log in with the customer’s password. That makes fraud much harder to spot.

There is also a business reason. Banks worry that smart agents could quickly move customers’ idle cash to accounts that pay higher interest elsewhere. So this is partly about security, and partly about control.

The bigger picture: banks under attack worldwide

The US story is part of a wider trend. In just the last month:

• South Korea

  Shinhan Bank

  Breaches hit seven financial firms. At Shinhan Bank, loan application data for about 25,000 people leaked. Attackers got in through loan-agent websites, not the banks’ main systems, and regulators are looking into whether they used AI.

• India

  State Bank of India

  The country’s largest bank is investigating after nearly 4,000 customer records showed up on the dark web.

• Bangladesh

  Janata Bank

  The bank’s website was attacked even after the bank took it offline as a precaution.

The pattern is clear. Attackers look for the weakest door: a partner’s website, an old system, a stolen password. AI helps them find those doors faster and cheaper than ever.

Why AI security is important

For years, AI security meant stopping a chatbot from saying something wrong. Now AI can act, and that changes everything.

• AI now has hands, not just a voice. A wrong answer is embarrassing. A wrong action, like sending money, deleting data or sharing private files, causes real damage.
• Mistakes happen at machine speed. A human might make one bad payment. A broken or hijacked agent can make thousands before anyone notices.
• Trust is the whole business. Banks, hospitals and governments run on trust. One AI-driven breach can undo years of reputation.
• The rules are still being written. Bills in the US Congress on AI agent accountability are still waiting. Until the law catches up, companies must protect themselves.
• Attackers use AI too. Defenders cannot use yesterday’s tools against tomorrow’s threats.

The more power we give AI, the more we must control, watch and limit it.

What you can do

Final thoughts

AI agents are coming, and they will make life easier. But the banks’ decision shows that convenience without control is a risk nobody wants to carry.

The winners will not be the companies that use AI fastest. They will be the ones that use it safely. AI security is no longer a feature to add later. It is the foundation everything else stands on.

What do you think: would you let an AI agent manage your money?

Tell me on LinkedIn or by email. You will find both links at the end of this post.