Major US banks, such as Bank of America and JPMorgan, have detected activity on some accounts carried out by third-party artificial intelligence (AI) agents that were not following instructions from their customers.
The increasing presence of AI agents in digital services poses new risks that banking entities have begun to identify, as they are not able to detect whether the entry into their clients’ accounts corresponds to a person or a ‘bot’.
This problem worries the main American banks, such as Morgan Stanley, Bank of America, JPMorgan, City and Wells Fargo, among others, according to reports from Axios, which cites sources close to the matter.
In some cases, the activity of agents authorized by the clients themselves, who have access to the login credentials, has been detected. In others, they have identified that agents take action regardless of the instructions they have received from users.
The inability to discern who is accessing the account has led financial institutions to begin blocking access by third-party AI agents and amending their terms accordingly.
According to the aforementioned media, executives understand the need to find a middle ground, which does not reach a complete ban on agents given the growing tendency to use this technology.
One of the concerns lies in being able to distinguish human activity from automated activity and, especially, that of authorized ‘bots’ from those that may act fraudulently.
The sources consulted have also expressed concern about the pace at which AI agent technology is developing, with its ability to identify new vulnerabilities, and the need to have adequate measures to deal with worst-case scenarios.