Introduction

Artificial intelligence agents are increasingly capable of logging into bank accounts, reviewing spending habits, and even executing payments on a customer's behalf. While these tools promise convenience, a growing number of major financial institutions are drawing a hard line: if an AI assistant moves money or exposes login details, the account holder—not the bank—bears the consequences.

What Happened

Citigroup recently revised its Online User Agreement to warn customers that sharing banking credentials with third-party AI tools comes with significant risk. According to a bank representative, customers who authorize these agents to access their accounts are responsible for any activity carried out through them, including transactions initiated by mistake.

The policy goes further. Citi also stated that if an AI tool leaks a customer's credentials or suffers a security breach, the customer remains liable for whatever attackers do with the compromised account.

Citi is not acting alone. JPMorgan Chase and Bank of America, the two largest banks in the United States, introduced comparable terms last month. PNC, ranked as the eighth-largest bank, updated its agreements this week. Citizens Financial Group has also notified customers by email that they are on the hook for actions taken by their AI assistants.

Why This Matters

The shift places a new burden on consumers who are already navigating the early stages of AI adoption. Handing over bank login information to an autonomous agent requires a level of trust that many users may not be ready to give, especially when the bank will not step in to reverse unauthorized transfers or cover losses from a hack.

At the same time, banks see AI-driven finance as both an opportunity and a competitive threat. Tools that help shoppers avoid fees or move money out of low-yield accounts could erode traditional banking revenue. Some institutions are developing their own AI assistants, and Citi's terms indicate that customers using bank-provided agents may not face the same liability as those using outside software.

Key Takeaways

  • Citi, JPMorgan Chase, Bank of America, PNC, and Citizens have all updated policies to place liability on customers for third-party AI agent activity.
  • Customers may be responsible for intentional or accidental transactions initiated by AI tools they authorize.
  • If an AI agent leaks credentials or is hacked, the account holder could still be liable for resulting losses.
  • Bank-branded AI agents may carry different terms than third-party tools.
  • Anthropic recently launched a read-only personal finance tool that cannot move or invest money, highlighting a more cautious approach.

Conclusion

As AI agents move from experimental tools to everyday financial assistants, the responsibility gap between what these systems can do and who pays when things go wrong is becoming clearer. For now, anyone considering connecting an AI agent to their bank account should read the updated terms carefully—and assume that, in the eyes of most major banks, they are on their own.