AI assistants are no longer just answering questions. They are booking flights, managing calendars, and increasingly, being handed access to bank accounts. Bank of America wants customers to understand exactly what that means if something goes wrong.
The bank sent customers an email titled “Protecting your finances in the Al World,” warning them not to share their online banking usernames or password vaults with any outside service, explicitly including AI-enabled tools or agents acting on their behalf. A Bank of America spokesperson confirmed the email is authentic.
What the Warning Actually Says
Bank of America’s service terms place the responsibility squarely on customers who voluntarily share their credentials.
The warning carries real consequences. Under the bank’s existing service agreement, customers who voluntarily share credentials are responsible for activity resulting from that access, even if an AI agent acts beyond what the customer intended. That responsibility holds until the bank is notified and the credentials are blocked or replaced.
The bank’s online banking agreement states that customers generally may not authorize a third party to use their credentials on their behalf, with narrow exceptions such as legal guardianship or power of attorney. Beyond the liability question, Bank of America flags a data-control problem: when you hand over your login to a third-party service, you typically cannot control how much of your account information that service can see. The third party may store those credentials, and if that service suffers a breach, your banking password could be exposed along with it. Apps have also been known to be secretly tracking users in ways they never anticipated.
Why AI Agents Make This Worse
AI agents introduce a specific risk layer that credential-based access to traditional third-party apps does not fully capture.
A third-party budgeting app logging into your account is one thing. An AI agent operating with the same credentials is another. The agent can potentially initiate transactions you did not intend, misinterpret an instruction, or execute a move that is difficult to reverse.
Think of it as the difference between giving someone your debit card to buy groceries and handing that same card to an autonomous system told to “handle your finances.” The scope of what happens next is much harder to predict. A compromised AI service could expose stored banking credentials to attackers. Automated account activity may also trigger the bank’s own fraud controls, though Bank of America has not confirmed a specific policy that locks accounts whenever agent activity is detected.
What You Should Actually Do
A safer connection option already exists, and if you have shared your credentials, the bank’s guidance is clear.
If you want to connect a financial tool to your Bank of America account, use the bank’s official consent-based connection process rather than entering your credentials directly into a third-party app or AI service. Through this method, you can review which services have access to your account and revoke certain permissions through online or mobile banking.
If you have already shared your banking password with an AI tool or another outside service, Bank of America recommends changing your password as the safest immediate step. To stay safe, the broader industry, including banks, payment networks, and AI developers, is still working through agent authentication, transaction limits, and liability frameworks for agentic payments. No comprehensive standard is settled yet.
Your AI assistant may be capable of taking action on your behalf. Whether it is capable of making the right call with your money is a separate question entirely.
