# Wells Fargo's New Terms Make Customers Responsible for Their AI Agents

_Wells Fargo's online banking agreement, in effect from November 24, 2026, adds a section on AI agents. A customer who gives an agent a login answers for what the agent does, including actions it was never asked to take, while the bank may make customers register agents and may limit or block them._

By Johannes Keienburg, CEO & Founder  
Published: 2026-10-27  
Source: https://www.cakewalk.security/blog/wells-fargo-ai-agent-terms-customer-responsible

---

## Wells Fargo Will Treat What an AI Agent Does as the Customer's Own Act

From November 24, 2026, Wells Fargo will treat what a customer's AI agent does in online banking as the customer's own instruction, unless the law or a regulation provides otherwise. The bank added the rule to its Online Access Agreement, the contract behind its online banking, as a new section on AI agents that the version in force since May 21 lacked. The agreement defines an AI agent as any software that acts on behalf of a person, whether autonomously or not. The rule applies once a customer gives an agent their login or authorizes it in any other way. The section settles in contract a question any service that lets agents act for people has to answer: who is responsible when an agent goes beyond what its owner asked for.

> Any instructions, actions, or transactions initiated by the AI Agent are deemed to have been initiated and authorized directly by you, without further verification, and we have no obligation to investigate the validity or scope of such authority.

_Wells Fargo, Online Access Agreement, Section 15(d), version effective November 24, 2026_

The customer is "solely responsible" for monitoring the agent and for keeping it "within the scope of authority you have granted to it or in your best interest". That responsibility covers unintended actions, actions caused by "manipulated inputs" and an agent "exceeding the authority granted by you", again "unless otherwise provided by law or regulation". The bank states that it does "not monitor, screen, or review AI Agent activities". An existing clause already made customers answer for transactions by people they let into the account, including ones "you did not want or intend". The new version adds those people's AI agents to it.

Ending an agent's authority is the customer's job as well. The customer must revoke an agent's authorization, keep evidence of having done so and show that evidence to the bank on request. Until the bank has been notified, the customer stays responsible for everything the agent does, to the extent the law allows.

## Wells Fargo Keeps the Right to Limit or Block Agents

Section 15(g) reserves the bank's own rights over agents, among them to:

> Limit, prohibit, prevent, suspend, or terminate access by any AI Agent to the Service or any Eligible Account, including by restricting certain actions by AI Agents.

_Wells Fargo, Online Access Agreement, Section 15(g), version effective November 24, 2026_

The same section lets the bank require customers to [register their agents](/glossary/kya-know-your-agent) and change or shut down the programming interfaces agents use. The agreement does not say which agents the bank allows today. Wells Fargo's overview of the changes describes the section as the customer's responsibilities "if and when we permit the use of such agents". An [Internet Engineering Task Force (IETF) draft on agent authorization](/blog/ietf-ai-agent-authorization-draft), adopted by a working group in September, asks the receiving service to [decide each request](/glossary/ai-agent-access-control) using both the agent's identifier and that of the person it acts for.

> A bank that does not check what an agent does can only make the customer answer for it. Inside a company, the same blur appears when an agent works with an employee's own credentials. Cakewalk's MCP Gateway, which carries agents' tool calls to a company's apps, ties each agent session to the person it acts for. It checks every governed call against policy before the call runs, which lets a company keep the agent's access at the least it needs. One record of each call allowed, denied or escalated then shows the company what the agent did through the gateway.

_Comment from Gil Röder, CPO & Co-Founder_

Source: Wells Fargo, [Online Access Agreement](https://www.wellsfargo.com/online-banking/online-access-agreement/upcoming/), version effective November 24, 2026.