AI agents can now log into bank accounts, schedule transfers, and pay bills on a customer’s behalf. Wells Fargo’s revised Online Access Agreement, set to take effect November 24, 2026, makes the bank’s reported contractual position clear: anything that agent does is treated as if the customer did it directly.
Reported Section 15 of the agreement states that any instruction, action, or transaction initiated by an AI agent is deemed to have been initiated and authorized directly by the customer, without further verification from the bank.
“At Your Own Risk” Is Doing a Lot of Work Here
The agreement reportedly places the burden of monitoring and controlling an AI agent entirely on the customer.
The agreement reportedly describes AI agent use as “AT YOUR OWN RISK,” a phrase drawn from a contractual provision that does not, by itself, establish absolute legal liability. That language means the customer is responsible for monitoring the agent and keeping it within the authority granted to it.
The reported terms extend that responsibility further. Unintended actions, activity triggered by incomplete or manipulated inputs, and cases where the agent exceeds its granted authority all fall on the customer, subject to applicable law or regulation. Wells Fargo simultaneously disclaims any obligation to monitor or review what an agent does inside an account, while reserving the right to block or terminate agent access at its discretion.
Disabling the App May Not Be Enough
The practical trap is that turning off an agent and notifying the bank are two different actions.
Under the reported terms, liability for an agent’s activity continues until Wells Fargo is notified that authorization has been revoked, not simply until a customer switches the agent off. If a customer disables an AI tool through the app itself but never notifies the bank directly, the contractual clock may still be running.
Customers are reportedly required to revoke access properly and retain evidence of that revocation, including timestamps, screenshots, and confirmation records. That documentation must be provided to Wells Fargo upon request.
Other Banks Are Taking a Different Approach
JPMorgan Chase, Citigroup, and Charles Schwab have each drawn a harder line on third-party agent access.
JPMorgan Chase has reportedly moved toward blocking outside agents from directly accessing customer accounts, while permitting some read-only open-banking access. Citigroup reportedly bars third-party agents from logging in or initiating transactions at all, and Charles Schwab reportedly blocks unauthorized agents attempting to move client money.
Wells Fargo’s model differs by conditionally permitting agent access and then allocating the operational risk to the customer.
The Contract Doesn’t Settle the Legal Question
Existing consumer-protection rules may constrain how much of this risk a bank can contractually reassign.
The reported provisions include a consistent qualifier: “subject to applicable law or regulation.” That phrase leaves significant room. Existing consumer-protection rules governing unauthorized electronic fund transfers may constrain how much of this risk a bank can contractually push onto a customer.
The unresolved question is narrower than the contracts suggest: a customer granted access voluntarily, but did not authorize a specific transaction the agent later executed. The available reporting does not identify a clear controlling answer from courts or regulators, and the Wells Fargo agreement does not resolve it either.
What to Do Before You Connect an Agent to Your Account
These steps are editorial guidance to reduce exposure, not requirements that Wells Fargo legally mandates.
Before connecting any AI agent to a financial account, customers should confirm exactly what permissions the agent receives, specifically whether it can initiate transfers or payments. Any revocation should be recorded with a timestamp, with screenshots or confirmation emails retained. Customers should notify the bank directly when revoking access rather than relying solely on disabling the agent through its own interface. Checking a bank’s current terms to understand how agent-initiated transactions are classified is also advisable.
A Gap Between Marketing and Enforceable Rules
Financial institutions are moving faster than the law to define who bears the cost of an AI mistake.
The real story is not that banks have definitively made customers liable for every AI error. Financial institutions are beginning to allocate that risk contractually while the legal framework remains unsettled.
Anyone connecting an AI agent to a bank account is currently operating between a marketing promise and an enforceable rule. Understanding that gap matters before the agent starts moving money.





























