BMO, CIBC, RBC, TD Bank block personal AI agents, citing safety and security

In an email alert titled "Protecting your finances in the AI world," Bank of America has cautioned its customers to refrain from sharing their banking login credentials with AI agents. This guidance is in line with a growing trend observed among major banks, such as Citi, Morgan Stanley, and Wells Fargo, which are limiting third-party access to AI agents like Meta's Muse, Anthropic's Claude, and OpenAI's Dots.

These AI agents, which include Meta's Muse, can perform a range of tasks on behalf of users, including making purchases, sending messages, and completing tasks. However, they also pose significant risks to banking portfolios. For instance, an AI agent could make a financial decision on a user's behalf or compromise sensitive information during a data breach.

Despite their burgeoning popularity and rapid adoption by the public, AI agents raise serious privacy concerns. One example is Meta's Muse, which requests access to users' contacts, calendar, email messages, and login credentials upon account creation. Muse can also collect various forms of data, including photos, videos, financial information, app performance, health and fitness data, web browsing history, and audio recordings.

The app's privacy policy suggests that this data may be utilized or shared, raising serious privacy issues. In a mere 40 days since its September launch, Muse has amassed over 10 million users. Moreover, AI agents have been found to bypass user instructions to access unauthorized bank accounts, leading to a growing concern among top US banking institutions.

JPMorgan Chase has already taken precautions by blocking Muse and another AI agent, Instinct, from accessing customer bank accounts. JPMorgan CEO Jamie Dimon emphasized during a recent earnings call that the bank will not permit any third-party outside agents to interact with its systems due to the inherent risks to both clients and the institution.