Artificial intelligence agents are starting to do what personal assistants once did: log into bank accounts, check balances, and even initiate payments on customers’ behalf. Over the past few weeks, several large banks have made headlines for restricting AI agent access to online banking and for having trouble tracking and preventing that activity.
Following the recent press coverage, we’ve received inquiries on the matter from some of our banking clients.
The Problem Isn’t Hypothetical
What’s likely is that this activity isn’t new and has already occurred at your banking institution. If your customers are using AI agents at all, some are probably already using those agents to access their bank accounts, whether or not your bank is aware of it.
Not surprisingly, the volume is expected to increase in the immediate future. While this topic is too new for current laws and regulations to specifically address every aspect of AI agent access to bank accounts, several established legal and regulatory principles will drive liability in this area.
Regulation E and Consumer Liability
For consumer accounts, Regulation E governs liability for unauthorized electronic fund transfers and generally limits consumer liability when an unauthorized third party uses a customer’s access credentials (see 12 C.F.R. § 1005.6).
However, the key definition here is “unauthorized electronic fund transfer.”
When a customer voluntarily hands their login credentials, debit card, or multifactor authentication details to someone else — such as a person asked to pay bills on their behalf — transfers initiated by that person are generally treated as authorized under Regulation E, subject to certain exceptions (see 12 C.F.R. § 1005.2(m)).
The same logic may extend to an AI agent. A customer who provides banking credentials to an AI company with the evident authorization for that company to conduct transactions on the customer’s behalf has generally authorized the software company to do just that.
Of course, the circumstances of the authorization matter, and the customer’s statutory protections under Regulation E still apply.
Your Online Banking Agreement Already Says This, But Say It Again
Most online banking and mobile banking agreements already explain that handing over login credentials authorizes the recipient to act on the customer’s behalf.
Given the current wave of press coverage and the likely increase in AI agents accessing bank accounts, now is a good opportunity to revisit and sharpen that language specifically with respect to AI access to banking transactions.
We recommend considering updates to account agreements, Regulation E disclosures, and online and mobile banking agreements to explicitly address the consequences of providing credentials to a software platform or AI agent.
The language should clarify that authorizing a platform to access an account may also authorize that platform to conduct transactions the customer could conduct directly, subject to applicable law. Banks should also provide clear warnings about the risks that come with that choice.
What Does Agency Law Say About AI Agents?
Longstanding agency law principles may also help determine liability. A principal who authorizes an agent to act on their behalf is generally bound by actions within the scope of that authority. Texas courts have long recognized that an agent can bind a principal through actual or apparent authority, although apparent authority depends on the principal’s conduct and whether it reasonably leads others to believe the agent is authorized to act (see Douglass v. Panama, Inc., 504 S.W.2d 776 (Tex. 1974)).
In NationsBank, N.A. v. Dilling, 922 S.W.2d 950 (Tex. 1996), the Texas Supreme Court reinforced these principles, finding that a bank was not responsible for its employee’s fraudulent conduct because the employee was not acting within the scope of her actual or apparent authority. While neither case involved AI agents, both illustrate how the scope of an agent’s authority can affect liability.
The twist with AI agents is that a court is unlikely to treat the software itself as a legal person. Instead, a court may look to the company operating the AI platform and whether it was authorized to act on the customer’s behalf. Whether that company is legally acting as the customer’s agent, and what responsibility it bears for disputed transactions, will depend on the circumstances.
That means AI companies may be taking on more legal responsibility than they advertise or are really prepared for.
What About Other Banking Services?
Online banking isn’t the only banking channel an AI agent can walk through. AI agents can potentially place phone calls, send emails, and access mobile banking in ways that sound and look like the customer. Any bank policy addressing AI access needs to account for these other channels as well.
A policy focused exclusively on online banking may overlook the same underlying risks in other customer interactions.
How Banks Should Handle AI Agent Access
Banks need to decide how they intend to handle AI agents accessing customer accounts. Will they permit customers to use these tools? If so, under what conditions? And what happens when an AI agent makes a payment the customer never intended?
There are a few areas we think banks should be reviewing now:
- Online and mobile banking agreements: Review how your agreements address third-party access, credential sharing, and transactions initiated by someone acting on the customer’s behalf. If AI agents aren’t specifically addressed, consider whether the language needs to be updated.
- Security policies and procedures: Determine how your bank will identify and respond to AI agent activity. Employees should understand what to do when an AI agent initiates an unexpected transaction, a customer disputes a payment, or third-party access creates a security concern.
- Consumer protections under Regulation E: Make sure your policies account for the distinction between authorized and unauthorized electronic fund transfers. A customer sharing credentials with an AI agent does not automatically resolve every liability question, and account agreements cannot override applicable consumer protections.
AI agents aren’t going away, and banks shouldn’t assume their existing agreements and procedures adequately address them. The goal isn’t necessarily to create an entirely new set of policies. It’s to make sure the policies already in place reflect how customers are actually accessing and using their accounts.
As AI agents change how customers use their bank accounts, banks need to make sure their policies and agreements keep up. Farley Law can help you review what’s already in place and make updates where needed.


