Category: Industry Analysis | Reading time: 3 min
Major U.S. banks are telling customers in writing that whatever an AI agent does inside their account is the customer's responsibility. Business Insider reported on November 19 that it reviewed emails and in-app notices from Ally, Citi and Chase, and Wells Fargo has rewritten its online access agreement along the same lines.
What the banks are saying
Ally emailed customers earlier this month: "When you use an AI service, you're responsible for any actions that the AI takes." The email added that AI "can make mistakes, hallucinate, misunderstand your instructions, or follow them in ways you didn't intend, which could lead to significant loss."
Citi's notice says AI assistants can make mistakes, "and that means you can be liable for errors it makes." Chase warned in a November email that "authorized transactions may not be reversible if your information is misused." Every bank email Business Insider reviewed told customers not to share their username, password or one-time codes.
Wells Fargo goes further
Wells Fargo's upcoming online access agreement treats transactions made by an authorized representative or delegate as authorized unless the customer tells the bank otherwise, and makes customers responsible for AI agents "if and when we permit the use of such agents." American Banker reported on November 2 that Wells Fargo appears to be the first bank to require customers to show they have turned off an agent's access, or remain liable for what it does.
A Wells Fargo representative told Business Insider: "No technology should be trusted to act on a customer's behalf without strong security, clear customer authorization, and safeguards."
Why it matters
The banks' position follows months of agents showing up in customer accounts. Axios reported in October that large banks were detecting third-party agents operating inside accounts, and JPMorgan CEO Jamie Dimon said on the bank's Q3 call, "You don't want rogue agents, and there could be thousands of them." Retail is moving the same way: Target changed its terms in March so that purchases by a shopping agent count as authorized by the customer.
The effect is that liability for agent actions is being assigned by contract before any regulator has set a rule. For any business that lets an agent act on a financial account, the questions that decide who pays are now evidence questions: what the agent was authorized to do, what it actually did, and when its access was revoked. Without an independent, time-stamped record of each of those, a customer has little to show a bank after a disputed transaction.
Source: Business Insider, American Banker, Wells Fargo online access agreement (upcoming)