Bank of America has told customers that anyone who gives an AI agent their banking username and password is responsible for what that access does. The October 27, 2026 email, which a bank spokesperson confirmed to TechCrunch, addresses AI agent credentials directly and settles who carries the risk when an agent logs in as the customer.
Bank of America’s Email on AI Agent Credentials Puts Liability on the Customer
Bank of America’s email tells customers never to give their username and password to another person or an outside service, “including an AI-enabled tool or agent that offers to act on your behalf.” It adds that customers who share their credentials are responsible for activity resulting from that access. The message, headed “Protecting your finances in the AI World,” was posted to Bluesky on October 27 by Bryan Toporek, and TechCrunch obtained the bank’s confirmation that it is genuine. The same email says Bank of America continues to invest in AI for its own services. The bank draws its line between AI it builds into its products and outside agents that sign in with a customer’s password.
Bank Terms Already Treat Shared-Login Activity as the Customer’s Own
The liability position in Bank of America’s email matches language already written into bank account terms. U.S. Bank’s digital services agreement, which TechCrunch cited, says any activity performed with a customer’s login credentials is presumed authorized by that customer unless the customer reports otherwise. It says anyone given the login becomes authorized to act on the customer’s behalf, and that the bank is not responsible for managing those third-party relationships. A customer who wants to cut off shared access has to contact the bank to block digital access until new credentials are set. Under those terms an agent holding a password has no scope, no spending limit and no identity of its own. To the bank, it is the customer.
Delegated Access Lets an Agent Act Without the Password
The same U.S. Bank agreement describes the alternative to sharing a password. In most cases, it says, customers can share information by logging in directly with the bank and giving sharing instructions, which they can manage and revoke later. TechCrunch reports that banks, payment providers and processors are still working out how customers should give AI tools access, and Axios reported on October 22 that banks are contending with unauthorized agents reaching customer accounts. Some retailers and card networks have started building agent-specific routes. Walmart is giving each shopping agent its own profile with a spending cap, and American Express has a framework that checks purchase intent before an agent can buy.
Why This Matters for Engineering Leadership
If your product includes an agent that acts inside someone else’s account, Bank of America’s email describes your users’ exposure when that agent signs in with their password. The bank treats the agent’s actions as the customer’s, the customer is responsible for the results, and TechCrunch notes the bank may read the unfamiliar activity as an account takeover and lock the customer out. Three decisions follow. Do not ship an agent feature that asks users for banking passwords; build on an OAuth grant, an aggregator with a bank agreement, or the bank’s own sharing controls, and leave out any provider that offers none of those. Store any third-party tokens your agents hold with the same controls as your own production secrets, and give users a one-step way to revoke them. Teams building APIs that agents will call should read the email as evidence that customers hand over passwords when no scoped option exists. A delegated grant with limits and an audit trail tells you which actions came from an agent, which a shared password never will.
FAQ
What did Bank of America tell customers about AI agents?
In an October 27, 2026 email, it said customers should never give their username and password to another person or an outside service, including an AI tool or agent offering to act on their behalf.
Who is liable when an AI agent uses shared bank credentials?
Bank of America’s email says the customer is responsible for activity resulting from shared credentials. U.S. Bank’s terms presume any activity using a customer’s login was authorized by that customer.
Is the Bank of America AI agent email real?
Yes. A Bank of America spokesperson confirmed its authenticity to TechCrunch after it was posted to Bluesky.
Is there a safer way to give an AI agent access to a bank account?
U.S. Bank’s digital services agreement says customers can usually share data without sharing login credentials, by logging in directly and giving sharing instructions they can revoke.
Does Bank of America use AI itself?
Yes. The same email says the bank continues to invest in technologies including AI to make customers’ financial lives easier and more secure.