Smaller Banks Open Accounts to the AI Agents Big Banks Keep Out

Lead Bank lets the fintechs it serves open account numbers for their customers’ AI agents and enforces a spending ceiling on each account, while JPMorgan Chase blocks outside agents that log in with a customer’s password.

A company whose agents pay its bills now has to decide which bank holds that money. At JPMorgan, an agent working from a customer’s login leaves the risk with the customer, and Wells Fargo may treat an AI tool’s requests as the customer’s own. Lead writes the limits into the account itself, and Cross River confines each agent purchase to a single-use card.

How do the agent accounts work?

At Lead, the fintech sets which payees each agent account can reach and can cancel the account number once it has served its purpose. Column, a national bank, holds the wallets behind Natural, a startup whose software lets a company’s agents pay vendors by automated clearing house transfer, wire or real-time payment, each agent under its own limits.

Cross River takes the card route. The New Jersey bank stands behind Stripe’s card issuing for agents: Stripe’s Link wallet issues an agent a single-use virtual card for each purchase, so the agent never sees the customer’s own card.

Why are the big banks saying no?

JPMorgan is “not going to allow just any outside agents to come in inside the company, because it’s just too risky for clients and for JPMorgan,” CEO Jamie Dimon told analysts in October. CFO Jeremy Barnum said that if customers hand credentials to an agent that does something they don’t want, “that’s going to be on them.”

Wells Fargo’s updated online access agreement says requests through a third-party AI tool may be treated as the customer’s own, and reserves the right to block some tools. The terms also reportedly make customers keep proof that they revoked an agent’s access, and leave them responsible for the agent’s actions until Wells Fargo hears of the revocation.

Who carries the loss?

Lead Bank CEO Jackie Reses has reportedly proposed a split that runs the other way. Under it, a consumer would not be liable when an agent goes beyond the authority it was given. The bank would be liable when it controlled that authority and failed to enforce it, AI providers would answer for model defects, and merchants for knowingly pushing an agent past its limits.

Cross River argues for plain disclosures; its sample for a grocery agent tells consumers they will not be liable for an unauthorized purchase, which will be reversed within 24 hours. Consumers remain wary. In a Morning Consult survey for the American Bankers Association, 77% of U.S. adults said they were concerned about letting an AI agent move money, and 19% said an agent had already asked for their login or card numbers.

Wells Fargo’s new terms take hold Nov. 24 for every customer who keeps banking online with it.