Five banks shift liability for AI-agent activity to customers
Citi, JPMorgan, Bank of America, PNC and Citizens make customers bear losses from third-party AI agents, including transactions made after a system hack.
By Amara Osei
3 min read

What's News
- Citibank, the third-largest U.S. commercial bank, updated its customer terms this week.
- JPMorgan and Bank of America, the two largest U.S. banks, announced similar policies last month.
- PNC, the eighth-largest bank, changed its terms on the day TechCrunch reported.
- Citizens emailed customers this week, warning them about liability for AI-assistant activity.
- Anthropic launched its read-only personal-finance tool earlier this month; it cannot move, manage, or invest money.
By Amara Osei
Five U.S. banks have shifted liability for third-party AI-agent activity to customers; Citibank warns that account holders bear losses even if a tool is hacked.
TechCrunch reported that Citi, the third-largest commercial bank in the United States, updated its terms this week. Customers remain responsible for intentional or unintentional transactions made by third-party automated assistants.
Who bears the loss when an AI agent is hacked?
Citi’s language goes beyond responsibility for agent transactions. The bank says customers must also absorb losses if a third-party AI tool leaks their credentials, gets hacked, or gives attackers access to their accounts.
TechCrunch also reported that JPMorgan and Bank of America, the country’s two largest banks, announced similar policies last month. PNC, the eighth-largest bank, changed its terms that day.
A Citi spokesperson told TechCrunch: “As third-party AI tools become more widely available, we want customers to understand the risks associated with sharing banking credentials or authorizing those tools to access their accounts.” The spokesperson said Citi recently updated its Online User Agreement to reinforce customer responsibility for activity conducted through connected agents.
TechCrunch reported that banks broadly reject responsibility for intentional or unintentional transactions made by automated assistants. Customers bear both deliberate and accidental agent actions, including losses that follow security failures.
Citizens separately emailed customers this week, warning that they were responsible for anything their AI assistant did. The warning reached TechCrunch through an email seen by the publication.
TechCrunch described AI agents as ranging from frontier-lab models to purpose-built personal-finance tools. Users can give them banking credentials and access to other online financial portals.
They can analyze spending and investments, offer advice, handle transactions, and pay bills. TechCrunch said banks are wary of the unreliable AI actors now operating inside their systems.
TechCrunch said AI startups are looking for more revenue, and personal finance offers a route to add customer value and collect part of the fees generated by the sector. The approach gives AI companies a way to monetize access to financial accounts and the sector’s fee pool.
The report noted that consumers are already wary of sharing email or chat records with AI agents. Bank and brokerage accounts create a new trust challenge because the tools can handle money as well as analyze it.
Anthropic takes a narrower approach. The personal-finance tool it launched earlier this month is explicitly billed as “read-only” and “can’t move, manage, or invest your money.”
The distinction matters because Citi’s terms treat bank-provided agents differently. Customers can use AI agents supplied by Citi without worrying about additional liability, according to the updated agreement.
Banks also view efforts by AI tools to enter banking as a competitive threat, TechCrunch said. Markets briefly shuddered at the prospect of AI tools making it easier for consumers to avoid unnecessary fees or low-interest deposit accounts that pad bank margins.
If AI agents become an important way for consumers to make financial decisions, banks will fight to control that pathway, according to TechCrunch. The report has seen the same tension in car-hailing and food service, where AI labs promise to revolutionize customer engagement and service providers push back.
TechCrunch said how these relationships develop will determine which AI investments pay off. For now, the dividing line is control: customers can use Citi’s own agents without additional liability, while users of third-party tools remain responsible for agent errors and post-hack activity. Which platform controls the customer connection will determine whether banks or AI companies absorb the cost of mistakes and breaches.
Source: TechCrunch
More from Amara Osei
Show full bio
Senior reporter covering companies, deals and strategy in entrepreneurship, startups and venture funding at Business Bearings.
2,336 articles