Skip to content
← All stories

Identity · AI Security

Banks say customers are liable if AI agents get their accounts hacked

Major banks now warn customers that if an AI agent is hacked or leaks credentials, the customer, not the bank, is responsible for any losses.

Ross Grant2 min read

AI chip glowing inside a circuit board

The short version

Some banks now say that if you use an AI tool to manage your money and it gets hacked, you, not the bank, will have to cover any losses. This is a change from the usual policy where banks often refund unauthorised transactions. The banks are making these rules clear as more people try out AI-powered personal finance tools. If you use such a tool, check your bank’s terms and be careful about sharing your account details.

By the numbers

  • 5major US banks with new AI agent liability policiesTechCrunch
  • 1AI finance tool (Anthropic) with read-only access to reduce riskTechCrunch

What happened?

Several major US banks, including Citi, JPMorgan, Bank of America, PNC, and Citizens, have updated their policies to say that customers who use AI-powered personal finance agents are responsible for any losses if those agents are hacked or leak credentials. The banks are warning customers that they will not reimburse unauthorised transactions made by AI tools, according to TechCrunch. This shift comes as AI agents become more common in personal finance.

Who is affected

Customers at several major US banks, including Citi, JPMorgan, Bank of America, PNC, and Citizens, are now subject to new rules about using AI-powered personal finance agents. These agents can access bank accounts to analyse spending, offer advice, or even make payments and handle transactions. The banks have updated their user agreements to clarify that customers who give account access to third-party AI tools are responsible for anything those tools do, whether intentional or not. Citi told TechCrunch (opens in a new tab) it recently updated its Online User Agreement to make these risks clear.

What was taken

No specific breach or theft is reported in this story. Instead, the concern is about what could happen if an AI agent is compromised. If a third-party AI tool leaks a customer's credentials or is hacked, the customer will be responsible for any unauthorised transactions or losses from their account. This is a shift from the usual expectation that banks will reimburse customers for unauthorised transactions.

The response

Banks are warning customers directly. Citi, for example, has updated its terms and sent communications to make clear that customers are liable for activities conducted through AI agents they authorise. Citizens Bank emailed customers with a similar warning. JPMorgan, Bank of America, and PNC have also changed their policies to reflect this stance. The banks say these changes are needed as AI tools become more common and customers start using them to manage finances.

Some AI providers are responding by limiting what their tools can do. Anthropic, a major AI company, launched a personal finance tool that is strictly “read-only”, it cannot move or manage money. This is meant to reduce the risk if the AI agent is compromised.

What you should do

If you use an AI agent or personal finance tool that connects to your bank account, check your bank’s terms and conditions. Understand that you may be liable for any losses if the AI agent is hacked or misused. If you are considering using such tools, look for options that are "read-only" and cannot make transactions on your behalf. Never share your banking credentials with any tool or app unless you fully trust it and understand the risks.

Why it matters

This policy shift puts more risk on customers who use AI agents to manage their finances. If an AI tool is compromised, customers could lose money and have no way to recover it from the bank. It highlights the need for caution when sharing sensitive information with third-party apps.

Key points

  1. Major US banks now hold customers liable for losses if AI agents are hacked.
  2. Citi, JPMorgan, Bank of America, PNC, and Citizens have updated their terms.
  3. Customers may not be reimbursed for unauthorised transactions by AI tools.
  4. Some AI tools are limiting their features to reduce risk (e.g., read-only access).
  5. Customers should check their bank’s policies before using AI finance tools.

Sources

  1. TechCrunch

    Reference

The day's cyber news in 5 minutes.

Cyberend Daily is one short email with the stories that matter, written so anyone can follow.

Includes our free beginner roadmap · No spam · Unsubscribe anytime