Bank of America has told its customers to stop sharing their passwords with third-party AI agents, a direct warning that lands as autonomous software tools increasingly request login credentials to act on a user's behalf. The bank's guidance reframes a convenience feature as a security exposure.

What You Need to Know

AI agents are software programs that can log into accounts and complete tasks without a human clicking through each step. When a customer hands over a password, the agent, and any vendor behind it, gains the same access as the account holder. Banks generally treat shared credentials as a breach of their terms, which can complicate fraud claims. The warning signals that financial institutions are drawing a line between approved integrations and informal password sharing.

The Credential Handoff Problem

The warning targets a specific behavior: customers giving an AI agent their banking username and password so the tool can check balances, move money or pay bills. That handoff bypasses the bank's own authentication controls.

Once a password leaves the customer's hands, the bank can no longer verify who is actually logging in. The agent vendor, its cloud provider and any attacker who compromises either one inherit full account access.

Bank of America's position is that customers should keep credentials private and use official channels instead. The bank has not framed the warning as a product launch, but as a security boundary.

Why AI Agents Are Asking for Passwords at All

Most consumer AI agents were not built with banking APIs. They operate by driving a browser or app the way a person would, which means they need the same login a person uses.

That design choice pushes the security burden onto the customer. A properly integrated tool would use token-based access with limited permissions, so a breach would not expose the full account.

The gap between those two models is where the risk sits. Password sharing is the low-cost path for vendors and the high-risk path for users.

  • Credential exposure: A shared password gives the agent and its operator the same access as the customer.
  • Terms violations: Many banks treat password sharing as a breach of account agreements.
  • Fraud claim friction: Disputed transactions can become harder to resolve when credentials were voluntarily shared.
  • Third-party risk: A breach at the agent vendor can cascade into the customer's bank account.

Why This Matters

The warning sets a precedent that other banks are likely to follow. Financial institutions have spent years tightening authentication, and password sharing with an unvetted third party undercuts that work.

Customers who rely on AI agents for bill pay or balance checks face a practical choice. They can keep using the tools and accept the risk, or move to official bank integrations that limit what an agent can see and do.

Agent vendors, meanwhile, face pressure to build proper authorization instead of asking for passwords. The banks hold the leverage here, because they control the accounts and the fraud rules.

The longer-term effect is a split market. Tools that adopt secure access will be tolerated; tools that demand raw credentials will be pushed out of banking entirely.

What Customers Should Do Differently

The safest path is to stop entering bank credentials into any third-party agent. Customers should check whether their bank offers an official connection before granting access.

Anyone who has already shared a password should change it and review recent account activity. Enabling alerts for logins and transactions adds a second layer of visibility.

Bank of America's message is blunt, and it is likely to be repeated across the industry. Convenience, in this case, comes with a credential risk that banks are no longer willing to absorb quietly.