Several U.S. banks have recently started changing their terms of service to prohibit or restrict third-party AI agents from logging directly into users’ bank and credit card accounts. The banks affected include major institutions such as American Express, Chase, Bank of America, and Citi. Brokerages including Morgan Stanley and Schwab have taken similar steps.
The AI agents discussed here are not ordinary ChatGPT chatbots. They are AI tools that can automatically open webpages, log in to accounts, look up information, and even transfer money or make trades (I guess too many people have been using Muse, hahaha). Banks are mainly concerned that after users give an AI their usernames and passwords, it might take actions on its own without clear instructions.
The banks’ policies vary somewhat:
- American Express: Explicitly prohibits sharing account login information with third-party AI agents and prohibits unauthorized automated access. If a user violates these rules, AMEX may restrict account access, transactions, rewards, and other account features.
- Chase: New terms effective October 27 prohibit third-party AI from using a user’s username and password to log in. However, authorized access to account data through a bank-approved API is still allowed.
- Citi, Bank of America: Both have also started restricting third-party AI from logging directly into accounts and moving funds. Citi still allows authorized read-only data access through designated service providers.
- Wells Fargo: New terms effective November 24 are especially noteworthy. If a user authorizes an AI agent to operate an account, transactions the AI makes—even ones the user did not intend—may be treated as user-authorized actions. The user may be held responsible, subject to applicable legal protections.
Terms and conditions:
In my view, banks already offer good protection against credit card fraud. If a charge wasn’t made by the cardholder, it can generally be removed. But now that AI is involved, if something goes wrong, who is responsible—the user or the AI? Should the bank classify it as fraud or an unauthorized charge? It gets a little awkward. Also, given what AI can do today, it’s possible it could find bugs on a bank’s website. If everyone starts automatically exploiting rewards and deals, the banks won’t be able to handle it.