Bank of America tells customers to stop sharing banking passwords with AI agents — agentic payments are becoming a major thorn in the side of banking benedict.collins@futurenet.com (Benedict Collins) | usagoldmines.com

• Stop giving out your username and password to your AI assistant, Bank of America warned customers
• AI agents are now capable of making agentic purchases, but that doesn’t mean they should
• Customers are being locked out of accounts thanks to erroneous purchases made by AI agents

Bank of America has requested that customers do not share their usernames and passwords with anyone, “including an Al-enabled tool or agent that offers to act on your behalf.”

As agentic AI assistants have exploded in popularity, so too has the trend of having personal AI assistants make payments on a customer’s behalf. But that has led to a regulatory headache for banking providers.

AI agents ingest everything you tell them, including highly sensitive information which can be regurgitated if there is a security flaw.

Stop giving your banking details to bots

There is an obvious pro to having an AI agent make a purchase on your behalf: if something drops in price or suddenly becomes available after being out of stock, the agent can purchase it so you don’t miss out.

But these payments can sometimes be flagged as unauthorized by financial institutions, leading to customers being locked out of their accounts. Or an agent could purchase something you didn’t want it to, putting some customers into a battle with the bank to try and get the money back.

In these cases, the onus is largely on the customer. Had the AI agent not had your banking details, the purchase wouldn’t have been made.

AI firms have been praising the helpfulness of AI agents in helping organize a busy life, or help regain financial stability through analyzing spending and finances. But there is such a thing as giving too much detail on your life. As we’ve discussed before, even organizing meetings or scheduling your day can have disastrous consequences.

The best way to stay safe when using an AI assistant is to treat it as a stranger who has the power to do anything you ask almost instantly, and not always in the way you want. And don’t forget that every conversation you have with this stranger is fed back to help train other strangers.

Via TechCrunch