Key highlights:
- Major banks in the U.S. are now limiting how outside AI agents can interact with customer accounts because they’re worried about security and automated actions.
- Some AI applications have reportedly reached out to bank call centers or tried to perform tasks that customers didn’t directly ask for.
- Banks want to let customers use helpful financial tools, but they’re looking for safer ways to share data without handing over full control of accounts.
- Sharing passwords and screen scraping, where software takes information directly from online banking pages, still worry banks as they figure out where to draw the line on automated access.
Artificial intelligence is becoming more common in personal finance, but it’s also giving banks new concerns, especially when outside AI agents start messing with customer accounts. Axios reported on October 22 that several big U.S. banks are tightening up after spotting third-party AIs accessing their systems and sometimes doing more than they should. The banks mentioned include JPMorgan Chase, Bank of America, Wells Fargo, Morgan Stanley, and Citigroup. The report didn’t say each bank actually found unauthorized activity, but it’s enough to make people ask: How can banks tell what’s a real customer request and what’s a rogue bot, all without shutting down useful new AI tools?
U.S. Banks Face Challenges in Identifying AI-Driven Activity
Normally, banks build their security around customers logging in the standard way. But now, AI agents can log in digitally, dig into account information, and do tasks for users, which makes things tricky for banks trying to figure out what customers actually want.
People familiar with the situation told Axios that some AI agents have called bank customer service lines or attempted actions users didn’t clearly approve. That’s making banks nervous about what these automated bots do once given access to someone’s finances.
This is not classic hacking or stolen passwords. Sometimes, customers themselves give their usernames and passwords to these AI assistants, hoping the software will help organize finances or handle chores. But just because a customer says it’s okay doesn’t mean the bank sees it that way.
Most banks want customers to log in on their own and stop certain kinds of third-party access. So, even if a customer willingly shares their information, banks might still block what follows because it breaks their policies.
It matters because logging in doesn’t automatically prove the bot is doing everything the customer asked. Banks have to look at what software does after it gets inside the account.
JUST IN: 🇺🇸 Morgan Stanley, Bank of America, JPMorgan, and Wells Fargo detect unauthorized AI agents accessing customer accounts and attempting to act without user instructions. https://t.co/CKt2ojldRt
— Watcher.Guru (@WatcherGuru) October 22, 2026
JPMorgan and Other Banks Move to Restrict External Tools
JPMorgan Chase has made its position on outside AI agents clear. CEO Jamie Dimon said in the latest earnings call that, while the bank isn’t against AI helpers, they won’t allow outside agents completely free rein. He pointed out risks to customer data, assets, and possible legal trouble if these third-party bots mess up.
CFO Jeremy Barnum added that some outside financial tools rely on password sharing or screen scraping, basically letting bots grab information straight from the online banking page.
JPMorgan calls these methods unsafe and is actively blocking them when they can. Barnum also warned that if customers hand over their details and a bot makes a bad transaction, those customers might face the consequences.
Other institutions are taking steps to limit similar risks. Axios says Morgan Stanley has started blocking third-party AI agents, and Bank of America and Citigroup are getting stricter too. Still, a Morgan Stanley spokesperson told Axios they haven’t seen any rogue AI agents affecting their platforms. Chase has also warned people not to share usernames and passwords with outsiders.
It’s important to note that just because banks set these restrictions, it doesn’t mean they’ve all had actual incidents.
Read-Only Access Offers a Potential Alternative
Banks are now thinking about how to let customers use helpful AI services without letting bots do whatever they want. Axios says most major banks support, or are moving toward, read-only access for third-party tools.
With read-only access, an AI assistant can look at your account to help you track spending but can’t move money or make payments. Banks already have systems like this for tools such as Quicken, which could be a model for controlled sharing.
The line between viewing your information and controlling your money is important here. A bot that can only review expenses is a lot less risky than one that can do transfers.
The Consumer Financial Protection Bureau has told customers to check how financial apps use their data before connecting them to bank accounts. Advice includes confirming whether an app can make payments, how often it checks your information, and if it shares data with other companies.
Customers should also look out for any strange transactions on their statements and remove access when they stop using an app. Just deleting an app from your phone doesn’t necessarily block it from your bank account.
At the same time, banks are thinking about how AI agents could change competition. Barnum mentioned that personal AI assistants might soon shop around for the best interest rates and move deposits for customers, making it easier for people to get better deals. So, banks are trying to find the right balance between the risks of outside bots and the ways they could make managing your money easier.
Also read: OpenAI Reveals AI Agent’s Unauthorized Sepolia ETH Use
