FINAFinance faces

Banks warn customers they may bear losses from AI agents

AI agents that can shop and act online are prompting a warning from major banks: customers may be held responsible for transactions an agent makes using their financial information. Ally Bank, Chase and Citi have each cautioned users about the risks in emails, app notifications or website notices.

Biography OnlineNovember 19, 20262,662 reads0

Ally told customers that users are responsible for actions taken by an AI service, which can make mistakes, misunderstand instructions or act in unintended ways, potentially causing significant losses. Chase warned in a November email that authorized transactions may not be reversible if information is misused. Citi’s website and app say customers can be liable for errors made by AI assistants.

The alerts come as AI tools move beyond answering questions. Agents including Meta’s Muse, Instinct AI and OpenAI’s Dots can go online to complete tasks such as booking flights, reserving restaurant tables and buying products. Retailers have also set limits: Target changed its customer agreement in March to classify AI-agent shopping as transactions authorized by the customer, while Amazon blocked Muse agents from making purchases in September, citing safety and reliability concerns.

Banks are distinguishing authorized connections to AI services from sharing account credentials. Axios reported last month that major banks had detected third-party agents operating in customer accounts, sometimes with customers’ permission and login details. Banks whose messages were reviewed by Business Insider warned customers not to share usernames, passwords or security codes.

A Wells Fargo representative told Business Insider that no technology should act for customers without strong security, clear authorization and safeguards for their money and financial information. JPMorgan Chase CEO Jamie Dimon made a similar point on an October earnings call, saying users should understand the risks and that the bank would not allow just any outside agent into the company because it would be too risky for clients and JPMorgan. Apple’s latest Mac software beta also warns users considering giving an AI agent or app system-level access to a computer.

Comments (0)

Leave a comment

No comments yet. Be the first!