i
News
News · 2026-11-20

More banks say customers are liable for AI agents’ actions

Dataist Dataist

More banks are warning that customers are responsible for AI agents’ actions, even if the agents are hacked

Cover: More banks say customers are liable for AI agents’ actions

Citibank, the third-largest commercial bank in the US, has updated its terms of service for customers who give third-party AI tools access to their personal finances. Customers are now responsible for these assistants’ transactions, including if a tool makes a mistake or is hacked. JPMorgan and Bank of America have already announced similar rules, and PNC changed its terms today. Citizens has also warned customers that they are responsible.

Banks are disclaiming responsibility for AI agents’ actions

What customers are responsible for

Customers can give AI agents — from models built by leading labs to specialized financial services — login details for bank and other financial accounts. These tools analyze spending and investment decisions, offer recommendations, make transactions and pay bills.

Banks do not want to be held responsible for the actions of unreliable automated assistants. They warn that they are not required to reverse transactions an AI agent makes, whether intentionally or by mistake.

A Citi representative told TechCrunch that the bank had updated its online services agreement to clarify the risks of sharing banking information and granting access to accounts. Customers who allow third-party AI agents to work with their accounts are responsible for those tools’ actions.

Citi’s terms also cover data breaches: if an AI service exposes login credentials or is hacked, the customer is responsible for transactions attackers make through their account. This week, Citizens sent customers a similar warning by email.

JPMorgan and Bank of America, the country’s two largest banks, announced similar rules last month. PNC, the eighth-largest, changed its terms today.

Banks and financial AI services

For startups seeking new revenue streams, personal finance offers a way to provide customers with additional services and earn a share of the financial sector’s fees. But users may be reluctant to give AI agents access to their correspondence and chat histories, and access to bank and brokerage accounts raises the trust question more sharply.

Anthropic’s approach shows one way to limit the risks: the company’s financial tool, launched earlier this month, is presented as a read-only service. It cannot transfer, manage or invest money.

Banks also see financial AI services as competitors. Markets briefly reacted to the prospect that these tools could help customers avoid unnecessary fees and low-interest deposit accounts that boost banks’ margins. If AI agents become an important way to make financial decisions, banks will seek to control that channel of interaction. Citi’s terms specify separately that customers can use the bank’s own AI agents without taking on additional responsibility.

This is part of a broader business response to AI developers’ plans to reshape how people interact with service providers: companies say they should set the rules for access. Similar disputes have already arisen in ride-hailing and food delivery; their outcomes will affect which AI investments prove profitable.

Dataist’s take: Banks are drawing a clear line: a third-party agent may access an account, but the customer remains responsible for the consequences of its actions. Anthropic’s read-only approach offers one way to reduce this risk, but it does not resolve who controls the financial channel between customer and bank.

Daily AI news

Every day we pick what actually matters in AI and explain it plainly — no hype, no filler. Subscribe if you want to follow where the industry is going.

Only what matters — every day

Follow on X