More banks say customers using AI agents are on their own — even if they get hacked
Source Entity
Tim Fernholz

Citibank has updated its terms of service to disclaim liability for losses incurred by customers using third-party AI financial agents. This move highlights a growing industry trend where banks shift the burden of security risks onto consumers who grant automated tools access to their accounts.
The Shift in Liability: Banks vs. AI Agents
Citibank, the third-largest commercial bank in the United States, has recently updated its terms of service to explicitly distance itself from the security risks associated with third-party artificial intelligence tools. By stipulating that customers are solely responsible for any losses—including those resulting from hacking—when using these automated agents, the bank is setting a definitive boundary in the digital banking ecosystem. This decision reflects a broader institutional anxiety regarding the rapid integration of AI into personal finance management.
The Rise of Autonomous Financial Management
AI agents have moved beyond simple chatbots to become sophisticated personal finance managers. These tools, ranging from frontier large language models to niche financial applications, are designed to access user credentials to analyze spending habits, provide investment advice, and execute real-time transactions. While these features offer convenience, they create a massive attack surface. When a customer provides these agents with bank account access, they are effectively granting a third party the authority to act on their behalf, creating a complex web of digital authorization that banks are now refusing to underwrite.
Security Implications and Institutional Resistance
Financial institutions are increasingly concerned about the 'unreliable AI actors' that inhabit their systems. The core issue lies in the lack of transparency and control. If an AI agent makes an unauthorized transaction—whether due to a malicious hack, a coding error, or a flaw in the model’s logic—the bank faces significant difficulty in determining whether the action was authorized by the user or performed by a compromised bot. By shifting the liability to the user, Citibank is effectively signaling that the convenience of AI-driven finance comes with a transfer of risk that the bank is unwilling to absorb.
Historical Context and Financial Regulations
Historically, banking regulations like the Electronic Fund Transfer Act in the U.S. have provided consumers with protections against unauthorized transactions. However, these laws were written for a world of human-initiated digital banking, not for autonomous agents that act on behalf of the account holder. The current conflict highlights a legislative gap; as consumers delegate financial decision-making to algorithms, the legal definition of 'authorized use' becomes blurred. Banks are preemptively updating their terms of service to ensure they are not held liable for the erratic behaviors of third-party software.
Future Trends in FinTech
This move by Citibank is likely to trigger a domino effect across the financial sector. As AI integration grows, we can expect to see more banks instituting 'walled garden' approaches, where they may offer their own vetted AI tools while strictly prohibiting or penalizing the use of unverified third-party agents. For the consumer, this marks the end of the 'wild west' era of AI financial management, where users could freely link their accounts to any tool without considering the catastrophic implications of a security breach. Moving forward, the industry will likely demand standardized security protocols for AI agents before they are granted access to institutional financial data.