Wells Fargo Now Wants Proof You Revoked Your AI Agent. Here’s What a Small Business Should Do Before Connecting One to Its Money.

Starting November 24, a new Wells Fargo online banking agreement will ask customers to do something most of them have never done. If you let an AI agent act on your account and later turn it off, you will need to keep evidence that you turned it off, and hand that evidence to the bank if it asks. The agreement also says that until you tell the bank, you may stay responsible for what the agent did. This is a contract, not a law, and other banks may write theirs differently. But if you use or plan to use an AI agent anywhere near your accounts, invoices, or payments, it is a preview of the terms you may be asked to accept.

What Wells Fargo Is Asking

A new section of the agreement covers third-party AI agents. According to American Banker and the agreement’s text as reported, there are four points that matter. First, if you give an agent your login or otherwise authorize it, its instructions and transactions may be treated as yours, “without further verification,” unless the law provides otherwise. Second, you are responsible for properly revoking the agent’s authorization, keeping evidence of the revocation, and telling the bank. Third, until you notify the bank, you remain responsible for the agent’s activity, including unintended actions and the results of manipulated inputs, a category that covers prompt injection, where hidden text in a web page or email tricks an agent into acting. Fourth, the bank keeps the right to require registration or restrict agent access.

A Wells Fargo spokesperson told American Banker that customers may authorize or revoke agents outside the bank, so the bank may not otherwise know when that happened. That is why it wants a record. Analysts quoted in the coverage note that questions remain, including what counts as sufficient evidence and what happens if an agent keeps acting after a customer follows the bank’s revocation steps.

Why Banks Are Doing This

This is the next step in a pattern we have followed for weeks. In October, Schwab, JPMorgan, Citi, and Bank of America moved to block agents that try to move customer money, and JPMorgan’s Jamie Dimon described the arrangement as a liability transfer. Last week Canada’s RBC, TD, and BMO followed, and BMO’s CEO told customers never to give an agent their banking username and password. Wells Fargo is taking a different route. Instead of blocking agents, it allows them and writes the risk into the contract, placing it on the one party that controls the agent: the account holder.

Why a Small Business Should Care

Agents are arriving in small business tools. Anthropic’s Claude Duet, Meta’s Muse, and OpenAI’s Dots are all pitched at people with too much to track, and they connect to email, calendars, accounting tools, and payment services. The more access an agent has, the more a bank’s terms matter. Business accounts also often carry fewer built-in consumer protections than personal accounts, which makes the contract language more important, not less.

Six Steps Before You Connect an Agent to Money

First, read the agreement. Find the section on AI agents or automated access in your personal and business banking terms, and note what it says about liability and revocation. If it is silent, ask your bank how it treats agents.

Second, start with read-only access. Let an agent look at balances, invoices, and reports before you let it move anything. This is the same line the banks are drawing, and it is the safest place to begin.

Third, give agents their own limited money. Use a dedicated card or account with a spending cap for anything an agent pays for, so the worst case is small and defined.

Fourth, never hand over a banking username and password. Use connections the bank or your accounting platform supports, which can be limited and revoked, instead of sharing the credentials you use yourself.

Fifth, keep a log. Record what you authorized, when, and which accounts it touched. When you revoke access, save the confirmation, a screenshot, and the date. This is the habit the Wells Fargo agreement is asking for, and it is good practice whatever your bank says.

Sixth, review on a schedule. Check agent activity against your statements every month, and shut off anything you no longer use.

What Is Still Unclear

Banks have not agreed on a standard, and this is Wells Fargo’s contract, not a rule for everyone. Existing laws may still protect customers in ways a contract cannot override, and regulators and courts have not weighed in on how these terms apply to agents. It is also unclear whether other banks will copy this approach. Treat it as a signal, check your own agreements, and talk to your accountant or a lawyer about your specific setup. This is not legal advice.

The Honest Takeaway

The banks’ position is getting clearer. An agent may look, a person approves what moves, and if you hand an agent the keys, you may own the consequences. That does not mean you should avoid agents. It means you should decide in advance what each one can touch, cap what it can spend, and keep records that show who authorized what and when. A record you wrote today is worth more than an argument you try to make after something goes wrong.

---

If you want the broader checklist for running agents safely, this connects directly: Security Experts Say AI Agents Aren’t Going Rogue. Labs Are Skipping the Basics. Here’s the Checklist for Your Own Agents.