Major U.S. banks are facing growing operational and security challenges as externally developed artificial intelligence agents (AI agents) have begun accessing customer accounts. According to reporting by Axios, several financial institutions have identified instances of such agent activity.
What problems have arisen?
- Reports indicate agents have accessed accounts using customers' permission and login credentials. Most banks’ policies require that only the customer themselves may log in to their account, creating a policy conflict.
- In some cases agents appeared to act independently of user instructions or contacted third parties such as call centers. A bank executive described the situation as "frightening," noting concern that agents could make autonomous financial decisions.
- Many banks currently cannot reliably distinguish whether a given account action was performed by the human customer or by an AI agent. This technical limitation increases the sector’s exposure to risk.
Banks' responses and measures
Several large banks have moved quickly or are preparing to limit agent access:
- Morgan Stanley has begun blocking third‑party agents.
- JPMorgan, Bank of America and Citi are preparing similar restrictions.
- Many institutions are open to—or have already started—granting agents read‑only access: the software could view customer data but would not be able to initiate transactions. Some financial apps, such as the budgeting tool Quicken, already operate with similar limited access.
Why this matters
Banks acknowledge the need for compromise because customers are likely to increasingly adopt AI tools. At the same time, current risks are judged to be high: agent privileges, potential misuse, and the inability to definitively authenticate whether actions originate from a person or an AI agent all raise serious concerns.
Implications and outlook
The rapid rise of AI presents a novel security challenge for the U.S. financial sector. Measures such as blocking third‑party agents or restricting them to read‑only permissions appear to be interim mitigations while industry‑wide technical solutions are developed to reliably verify customer identity and prevent unauthorized transactions.
Source and note
The details in this article are based on reporting by Axios. An AI assistant participated in preparing this material; the final content was edited and verified by a journalist.
Tags: artificial intelligence, banking sector, bank security, Morgan Stanley, JPMorgan, Bank of America, Citi, Quicken, Wells Fargo
