People are giving AI agents their bank passwords, and banks are restricting that access, Axios reports. I understand why someone would want the help. Having an agent sort through transactions or help manage bills could save time and make everyday life easier.
I want that to work. But giving an agent my password leaves too much unanswered. What have I actually agreed to? Can it review my account, move money, or share information somewhere else?
The bank needs a way to recognize the agent and accept the connection. I need a way to choose what it can do. Knowing which agent showed up doesn’t answer whether it has permission to make a particular payment.
I’d start with access that lets it review information without changing anything. That’s useful on its own, though I’d still want to control which information it can see. If I later want it to pay a bill, I should be able to approve the amount and recipient before anything moves. That permission should expire when the job’s done, and I should be able to cancel it sooner.
Banks shouldn’t use security as an excuse to shut out every tool customers want to use. There’s worthwhile work here for banks and AI companies to figure out together. Axios’s follow-up reporting says banks hadn’t seen evidence of a hack affecting customer funds or accounts.
I’d like to see more useful connections built around clear permission. Asking AI for help shouldn’t require giving it every privilege my account has.