Wells Fargo’s updated Online Access Agreement takes effect on November 24, 2026, and its new Section 15 treats anything an AI agent does in a customer’s account as the customer’s own act, without further verification by the bank. The same section makes customers revoke an agent’s access themselves, keep evidence that they did, and hand that evidence over on request. These Wells Fargo AI agent terms are the most detailed written version of a position US banks took through October, when reporting showed them blocking third-party agents that log in with customers’ passwords.
Wells Fargo’s AI Agent Terms Shift Every Agent Error to the Customer
Section 15 defines an AI Agent broadly, as any software that takes “autonomous, semi-autonomous, or non-autonomous action” for someone, so a scripted budgeting tool falls under it as much as a large language model assistant. Section 15(b) says the bank does not provide, control, endorse, monitor or audit any agent, and that its own privacy and security policies do not apply to one. Section 15(d) says that if a customer gives an agent their login or otherwise authorizes it, the agent’s instructions are deemed authorized by the customer and the bank has no duty to check them. Section 15(e) then assigns the customer responsibility, unless law or regulation provides otherwise, for unintended actions, actions driven by manipulated inputs and actions that exceed the authority the customer granted. Wells Fargo’s summary of the changes says requests submitted through such a tool may be treated as if they came directly from the customer.
Revocation Now Needs a Paper Trail
The AI agent section is new: the agreement version in force since May 21, 2026 contains no AI agent provisions, and Section 15(f) is its most demanding clause. Customers must revoke an agent’s authorization “properly and timely,” take any further steps needed to stop its activity, retain evidence of the revocation and produce it when the bank asks. Until they notify Wells Fargo through the process in Section 17(d), they remain responsible for everything the agent does, to the extent the law permits. In practice that turns a settings toggle into a compliance artifact. A customer who deletes an agent from their phone and assumes it is gone carries the risk of every action the agent’s backend still takes with a stored session or password.
Banks Are Restricting Password-Based Agents While the Open Banking Rule Stalls
Wells Fargo is one bank in a broader pullback. Axios reported on October 22 that major US banks were finding unvetted third-party AI agents inside customer accounts, and Reuters reported the same day that large lenders have been blocking such agents from operating in accounts and making transactions. AI News, in its October 30 report, also cites JPMorgan Chase’s third-quarter earnings call, where the bank called credential sharing insecure and said customers who hand agents their logins own the outcome. The Consumer Financial Protection Bureau’s Section 1033 open banking rule was designed to replace password-based screen scraping with bank-provided interfaces, but AI News reports, citing law firm Cozen O’Connor, that a federal court has enjoined its enforcement while the Bureau reconsiders it. Section 15(g) of the Wells Fargo agreement anticipates the next step: the bank reserves the right to require customers to register their agents, to change or close agent access methods, and to block any agent outright.
Why This Matters for Engineering Leadership
If your team builds an agent that touches bank accounts, Section 15 moves your product’s failure modes onto your users’ balance sheets, and the bank has said in writing that it will not catch your mistakes. Three design changes follow. First, stop relying on stored customer passwords wherever the bank offers a scoped, revocable route, because a password cannot be limited to read-only and cannot be revoked without changing it everywhere. Second, build revocation as a feature with a receipt: when a user disconnects, destroy credentials and sessions server-side, stop queued actions, and give the user a timestamped record that names the account and the time access ended, since that record is what Wells Fargo can now ask them to produce. Third, log every instruction your agent sends with the user request that triggered it, so a disputed transaction can be traced to a prompt or a manipulated input. Platform teams inside banks should read the same clause from the other side: requiring agent registration under Section 15(g) is only workable if you can identify agents at login, and that capability needs to exist before the policy does.
FAQ
When do Wells Fargo’s new AI agent terms take effect?
The updated Online Access Agreement takes effect on November 24, 2026. Customers who keep using Wells Fargo online or mobile services after that date are bound by it.
What counts as an AI agent under Wells Fargo’s agreement?
Section 15(a) defines an AI Agent as any software or technology that takes autonomous, semi-autonomous or non-autonomous action on behalf of a person or entity.
Does Wells Fargo verify instructions sent by an AI agent?
No. Section 15(d) says agent instructions are deemed authorized by the customer without further verification, and the bank has no obligation to investigate them.
What does Wells Fargo require when a customer revokes an AI agent?
Section 15(f) says the customer must revoke the agent’s authorization in a timely way, keep evidence of the revocation and provide it on request. The customer stays responsible until notifying the bank.
Can Wells Fargo block AI agents?
Yes. Section 15(g) reserves the right to require agent registration, change API access, and restrict or terminate any agent’s access at any time.
What should teams building financial AI agents do before November 24?
Replace password-based access where a scoped, revocable alternative exists, and give users a timestamped revocation record they can hand to their bank.
