Bank of America Warns Customers Not to Share Banking Passwords With AI Agents

TL;DR

• Bank of America is reminding customers not to share banking usernames, passwords, or one-time passcodes with third-party AI assistants and autonomous agents, warning it can void fraud protections.
• The bank says customers remain responsible for transactions made by an AI agent acting on their behalf, even if the AI makes a mistake or is manipulated.
• The warning signals a growing clash between open agentic banking and traditional account security as AI assistants move to pay bills, transfer money, and shop.

A New Warning For The Age Of Agentic AI

Bank of America is drawing a hard line on AI access to your bank account: don't hand over your login.

In recent customer communications and updated security guidance, the nation's second-largest bank has warned customers against entering their online banking usernames, passwords, and one-time passcodes into third-party AI chatbots, browser assistants, and autonomous agents that offer to check balances, pay bills, move money, or analyze spending for you.

The message is simple - if you give an AI your credentials, you're giving it full control of your money.

Why Bank of America Is Speaking Out Now

The timing is no accident. Over the past year, a wave of AI assistants from OpenAI, Anthropic, Google, Perplexity and a host of startups have added agentic capabilities that can browse the web, fill out forms, and complete transactions on a user's behalf.

Many prompt users to log in to bank accounts, credit cards, brokerages, and payment apps so the agent can act for them. Some operate by taking over your browser or cloud computer and typing in the credentials you provide.

Bank of America says that model violates a core banking safety rule: never share your credentials with anyone - including an AI.

The bank points to a surge in phishing, credential stuffing, and AI-powered social engineering scams. Once credentials are pasted into a third-party AI tool, customers lose visibility into how they are stored, logged, shared with sub-processors, or potentially exposed in a data breach or prompt injection attack.

What Happens If Your AI Agent Sends Money

Perhaps the most important part of the warning is about liability.

Bank of America says customers are responsible for transactions they authorize - including transactions they authorize an AI agent to make for them. If you tell an agent to pay a bill, transfer funds, or buy something and it sends the wrong amount, pays the wrong person, duplicates a payment, or falls for a scam page, the bank may treat it as an authorized transaction you directed.

That distinction matters. Under Regulation E and most bank zero-liability policies, unauthorized transactions where someone steals your login and moves money without your permission are typically covered. Authorized transactions where you gave out your credentials and asked a tool to act for you are far harder to reverse, and may not be covered by fraud guarantees.

In short: if your AI assistant goes rogue, you may be on the hook.

Erica Vs. Outside Agents

The stance also highlights Bank of America's own AI strategy. The bank has spent years building Erica, its in-app virtual financial assistant, which now handles hundreds of millions of interactions for balance checks, spending insights, bill reminders, and fraud alerts.

Unlike outside agents, Erica operates inside Bank of America's authenticated app environment. Customers don't have to hand their password to a third party - they're already logged in securely, with biometric controls, device intelligence, and transaction limits.

Bank of America is effectively telling customers: use our AI inside our app, not someone else's AI with your bank password.

The bank has also endorsed secure data-sharing through APIs and open banking frameworks rather than credential-based screen scraping. Under that model, apps like budgeting tools connect via tokens with specific permissions and revocable access, rather than holding your full username and password.

What This Means For Agentic Banking

The fight over passwords gets to the heart of the future of agentic finance.

Tech companies want AI assistants to become a universal interface for money - able to log in anywhere and do anything you ask. Banks want that activity channeled through secure APIs, verified third parties, and their own apps, where they can enforce multifactor authentication, monitor for fraud, and limit liability.

Expect other large banks including JPMorgan Chase, Wells Fargo, and Capital One to issue similar warnings or quietly block agentic logins that look like bots. Regulators including the Consumer Financial Protection Bureau have already weighed in on open banking rules requiring secure, tokenized data sharing and clear liability when consumers authorize third parties.

For now, security experts say consumers should never paste banking passwords or one-time codes into an AI chatbot, browser plugin, or autonomous shopping agent.

How To Stay Safe While Using AI For Money

If you want to use AI to help with finances without risking your account, follow basic bank security hygiene:

Use your bank's official app and assistant first for balances, alerts, and bill management. If you use a third-party budgeting or AI tool, connect only through official OAuth or open banking prompts that do not ask for your full password. Never share one-time passcodes with an AI - those codes are meant only for you to complete a login you initiated. Turn on real-time transaction alerts, two-factor authentication, and card locks so you catch unauthorized activity fast.

Bank of America says if you have already shared credentials with an AI tool, change your password immediately, review recent transactions, and revoke access where possible.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.