Banks Have Had It Up to Here With Your AI Agent

Glass Epoch DeskOriginal digestSource: WIRED AI

AI-assisted digest, checked against the original source before publishing. How we work

Wells Fargo and a wave of US banks are rewriting terms to put AI agent risks—and liability—squarely on customers.
Banks Have Had It Up to Here With Your AI Agent
Photo: POURIA 🦋 · free stock license

Wells Fargo is rolling out an updated customer agreement on Tuesday that speaks directly to AI agents, and its language leaves little room for ambiguity: "Your use of any AI Agent is AT YOUR OWN RISK." The move is the most explicit example yet of a banking industry deciding that autonomous bots logging into customer accounts are its problem to fence off, not its product to support.

The Wells Fargo update lands amid a broader policy wave. In late October, JPMorgan Chase changed its terms to prohibit third-party agents from logging in with customer credentials, and Bank of America warned customers the same day against sharing login information with agents. Citibank and PNC have also revised their terms to include agent warnings. The pushback extends well beyond finance: Amazon blocked Meta's Muse agent from making purchases on its platform, and United announced that personal bots will be blocked from buying flights on a traveler's behalf.

Agents remain nascent technology, and the share of online transactions they handle is still minuscule, but a subset of power users has woven them into everyday tasks such as comparing couches, tracking price drops, and assembling grocery orders—often pairing agents with account credentials and broad permissions. "Consumers authorize goals, not every click," says Kacy Kim, an associate professor of marketing at Bryant University who studies consumer behavior and emerging tech. Forrester analyst Sucharita Kodali notes that most people don't even know what agents are, let alone use them, and sees the bank policies as the predictable behavior of institutions that are fundamentally risk-averse. "The banks see risks with everything," she says, from transaction security to automated withdrawals.

Consumer psychology research suggests the appetite for automation is narrow. Luca Cian, a professor at the University of Virginia who studies consumer psychology and AI, says people are very comfortable with AI buying routine household items like toilet paper but are far less willing to automate dynamic, big-ticket purchases that demand discretion. He offers a pointed hypothetical: an agent accidentally orders 10 roasted chickens. The retailer can claim it fulfilled a valid order, the AI builder can point to disclaimers, and the bank has now made clear it doesn't want the liability either.

What remains unsettled is where accountability lands when an agent's mistake is expensive rather than merely embarrassing. It's worth watching whether banks' warnings push builders toward bank-approved integrations instead of credential-based logins, whether platform blocks like Amazon's and United's spread across retail and travel, and whether early adopters—whom Kim expects not to be scared off—keep operating at their own risk while regulators or courts start assigning blame.

Key facts

  • • Wells Fargo's updated AI agent terms take effect Tuesday for millions of customers.
  • • JPMorgan Chase banned credential-based third-party agents in late October.
  • • Amazon blocked Meta's Muse from purchases; United blocked bot flight buying.
  • • Citibank and PNC also revised terms with AI agent warnings.
  • • Forrester says most consumers don't know what AI agents are.

Why it matters

Builders of consumer agents now face a permission wall at the account layer: banks and platforms are cutting off credential logins, so agent products will need bank- and retailer-approved integration paths to stay usable. Expect liability language to become a standard product surface—disclaimers help, but unresolved blame across retailer, model developer, and bank will shape which use cases (routi

Spotted an error? Send a correction.

Get the day's AI news in one email

The Daily Epoch · 60-second read · unsubscribe anytime

Related