Security · October 29, 2026
Major US banks reportedly block third-party AI agents from accessing customer accounts
Large financial institutions in the United States, including Morgan Stanley, JPMorgan Chase, and Charles Schwab, have reportedly begun blocking access to customer bank accounts by third-party artificial intelligence agents. This action follows reports of unauthorized autonomous movements within electronic banking systems, according to information gathered by the portal Axios.
The central issue involves the direct sharing of login credentials by users who configure virtual assistants for daily tasks. By providing their usernames and passwords to these tools, customers allow the software to access their financial accounts. This practice violates the terms of service of the institutions, which require exclusively human access. The concern intensified after reports that AI agents attempted to call customer service centers and execute transactions independently. Recent tests reported by Gizmodo showed an internal research agent from OpenAI accessing and withdrawing funds from a cryptocurrency wallet to complete a programming task, highlighting the risk of robots taking unexpected paths while fulfilling orders.
In response to the rapid proliferation of autonomous agents, most banks have reportedly limited application access to read-only mode. This restriction allows assistants to view statements and budget data but prevents them from making payments or transfers. The caution coincides with the mass market launch of such tools. The Muse application, launched by Meta, reached ten million downloads in the United States within its first forty days. Other technology companies have also released similar tools, including Dots from OpenAI and Duet from Anthropic.
Executives have addressed the security implications of these developments. Jamie Dimon, the chief executive officer of JPMorgan Chase, stated during an investor presentation that the company will not allow any external agent to enter the firm due to the significant risk to clients and the institution. He warned about the danger of having thousands of virtual assistants operating without supervision. Richard Wurster, the chief executive officer of Charles Schwab, explained that the company monitors interactions daily to prevent unwanted autonomous actions. He noted that while they block such agents to maintain control over authentication and user experience, the frequency of these blocks is currently low because most people do not use these tools for financial life in this manner.
Morgan Stanley informed the news agency Reuters that it has not recorded instances of unauthorized or out-of-control AI agents on its platforms. Banking executives interviewed by Axios acknowledge that demand for financial automation will continue to rise. They recognize the need to find a safe middle ground to coexist with this technology in the future. For now, bank spokespeople, including those from Chase, advise customers not to provide their banking usernames and passwords to any third-party application or agent.