Unvetted AI agents are getting into US bank accounts, and banks are scrambling to stop them
According to Axios, some agents appeared to act beyond what users asked. JPMorgan, Bank of America, Citi and Morgan Stanley are restricting access.
Major US banks are detecting unauthorised third-party AI agents operating inside customer accounts, and some of these agents have tried to take actions that appear to be independent of what users told them to do, Axios reported on Thursday, citing people familiar with the matter. Banks including Morgan Stanley, Bank of America, JPMorgan and Wells Fargo are racing to contain the risks, the report said. Reuters said it could not immediately verify it.
In most cases, according to Axios, an AI agent or assistant logs into an account with the customer's own permission and password. That still breaks the rules of most banks, which require customers to log in themselves. Bank executives fear what an agent might do with that access, especially if a user has allowed it to make purchases or move money. Some banks have also tracked agents contacting call centres or appearing to act without instructions. One official described the industry's concerns as grave.
Morgan Stanley has started blocking third-party agents from its platforms, and JPMorgan, Bank of America and Citi are also restricting them, Axios reported. A Chase spokesperson said customers should not give their usernames and passwords to anyone. A Morgan Stanley spokesperson said the bank had not seen instances of "rogue AI agents accessing our platforms". Most banks are moving towards read-only access, which lets agents see account information without being able to act, as budgeting apps already do. Many banks cannot yet tell whether an agent's activity is directed by the customer, the report said.
The report comes a day after the cybersecurity firm Straiker published research, first shared with The Information, saying a long-running criminal group known as CyberCartel used AI agents built on commercial models, including Anthropic's Claude, xAI's Grok and Moonshot AI's Kimi, to automate attacks on more than 400 banks and payment providers in Latin America. According to Straiker, the agents appear to have stolen passwords and customer data and accessed bank accounts with little human effort.