Are AI Agents Signing Into Your Customers’ Accounts?

When a customer hands their login credentials to an AI assistant, a bank can see which credentials were used but often cannot tell whether a person or an automated agent is behind the activity, or whether the customer actually approved what the agent did.

The Risk for Your Bank

When an AI agent signs in with a customer’s credentials, the bank inherits risks it cannot easily see:

• Disputed transactions and losses. If an agent moves money in a way the customer did not intend, a dispute is likely, and it is not yet settled who bears the loss under existing consumer protection rules such as Regulation E.
• Weakened authentication. Controls built to confirm a person is signing in lose their value when valid credentials are used by software, and a compromised or manipulated agent can look identical to a legitimate customer.
• Data privacy exposure. Account information that flows to an outside AI tool leaves the bank’s control, raising questions about how it is stored, shared, or used.
• Fraud monitoring blind spots. Automated activity can distort the behavioral patterns that fraud and BSA/AML monitoring depend on, producing false alerts or masking real ones.
• Reputational harm. When something goes wrong, customers rarely separate their AI tool from their bank.

Follow the example of larger institutions: set read-only access through approved data-sharing channels, with no ability for agents to move money. Community banks will face the same questions. We recommend preparing on two fronts.

Customer-Facing AI Agents

• Review your online banking terms. Confirm they clearly address credential sharing and automated access, and update them if they don’t.
• Talk with your digital banking vendor. If you use a 3^rd party, ask what tools are in place, or planned, to detect non-human login behavior and to support secure, read-only connections.
• Revisit fraud and dispute procedures. Consider how your team would handle a customer claim that an AI agent initiated a transaction they did not intend.
• Educate customers. Two-factor authentication and reminders not to share usernames and passwords, including with AI tools, goes a long way.

Vendor Use of AI

• Ask direct questions. Request that key vendors disclose where AI is used in the products and services they provide to you.
• Understand the data flow. Find out what bank and customer data their AI touches, where it is stored, and whether it is used to train models.
• Strengthen contracts. Seek advance notice of new AI features, clear limits on data use, and appropriate audit rights.
• Fold AI into your existing program. These reviews belong within your third-party risk management framework, not alongside it.

Banks don’t need to have every answer today. Keeping a risk-focus for access to customer accounts does require asking questions before moving forward.

If you would like to talk through how these developments may affect your institution, our team at GNA Advisors would be glad to help. Please don’t hesitate to reach out.

Recent Articles

• 
  Are AI Agents Signing Into Your Customers’ Accounts?
  Five of the nation’s largest banks, including JPMorgan Chase, Bank of America, Wells Fargo, Citi, and Morgan Stanley, have begun restricting third-party AI agents from logging into customer accounts.
• 
  Credit Union Times: When Deadlines Override Member Experience
  There is a pattern playing out inside credit unions across the country, and it is costing them more than they realize. Executive teams, under pressure from boards and annual planning cycles, push to get new technology live before Dec. 31.
• 
  Team Spotlight: Britney Purkey
  This week we are shining a spotlight on Britney Purkey, a servant leader who found her way to consulting by trusting her instincts and has never looked back.