Banks Crack Down on Third-Party AI Agents: New Terms Shift Liability to Users
Major banks like Wells Fargo and JPMorgan Chase are updating terms to disclaim liability for third-party AI agents, citing security risks while simultaneously developing their own internal AI tools.
S
··2 min readSciTech Editorial Team
Newsroom

Major financial institutions are moving swiftly to sever liability for third-party AI agents, marking a significant shift in the digital banking landscape. Wells Vanguard is set to enact updated customer agreements starting Tuesday, explicitly stating that the use of any AI agent is at the user's own risk. This decisive move follows similar actions by JPMorgan Chase, which banned third-party agents logging in with customer credentials in late October, and Bank of America, which issued warnings against sharing login information with autonomous bots. Citibank and PNC have also amended their terms to include specific warnings regarding these automated tools, signaling a coordinated industry response to emerging security threats.
The controversy centers on AI agents, which automate online tasks and execute clicks on behalf of users. While marketed as ultimate tools for shopping decisions—ranging from grocery orders to holiday gifts—these nascent technologies have stumbled through 2026. Early adopters have experienced both the efficiency of automation and public embarrassment when agents made critical errors, such as inadvertently posting sensitive bank details. Kacy Kim, an associate professor of marketing at Bryant University, notes that consumers authorize goals, not every individual click, yet they are increasingly being held accountable for the mistakes of their digital proxies.
Beyond the banking sector, AI agents are igniting turf wars across the web. Amazon recently blocked Meta’s Muse, one of the most widely used AI agents, from making purchases on its platform. Similarly, United Airlines announced that personal bots would be blocked from purchasing flights. Sucharita Kodali, a Forrester analyst, observes that while most people do not even know what agents are, banks are fundamentally risk-averse. She argues that proactive measures are necessary to prevent automated withdrawals or security breaches that could detrimentally affect account holders, even if current issues remain edge cases.
Consumer behavior studies suggest a nuanced relationship with AI automation. Luca Cian, a professor at the University of Virginia, indicates that users are comfortable with AI handling boring, low-stakes purchases like household supplies but remain hesitant to automate dynamic, big-ticket items. A hypothetical scenario of an agent accidentally buying ten roasted chickens highlights the complex liability chain: the retailer claims a valid order was placed, the AI builder cites disclaimers, and the bank refuses responsibility. However, certain credit card providers, such as American Express, are exploring limited protections for shoppers, potentially reimbursing customers for agentic mistakes if the original retailer refuses a refund.
Despite the crackdown on third-party tools, banks are heavily investing in first-party AI solutions. Wells Fargo’s generative AI tool has been used over a billion times, while Bank of America revamped its Erica Assist agent, and Citigroup launched an internal development platform centered on agents. Cian summarizes the strategy: banks are locking the door to outside agents while building their own. Nevertheless, Kodali remains skeptical about the necessity of agentic shopping, noting that purchasing on major platforms is already straightforward and that these tools may simply be an attempt by AI builders to generate affiliate revenue.



