Banks Hold Customers Liable for AI Agent Hacks
Citibank this week updated its Online User Agreement to tell customers that if they grant third-party AI agents access to banking credentials, they are responsible for activities conducted through those agents—including losses if the tools leak credentials or are hacked. A Citi spokesperson said the change clarifies the risks of sharing banking credentials or authorizing AI tools to access accounts. The move follows similar policies announced last month by JPMorgan Chase and Bank of America, the two largest U.S. banks; PNC changed its terms today, and Citizens emailed customers this week with a comparable warning, according to TechCrunch. The shift affects a growing category of AI agents—from frontier lab models to purpose-built personal-finance tools—that users can authorize to analyze spending and investments, offer advice, and even handle transactions and bill payments. Banks are signaling they will not reverse intentional or unintentional transactions made by automated assistants, and Citi’s terms extend that liability to breaches involving third-party tools. The stance adds a trust hurdle for AI startups seeking to enter personal finance, where access to bank and brokerage accounts raises cybersecurity concerns beyond those associated with sharing email or chat records. Some providers are already adjusting. Anthropic’s personal-finance tool launched this month is billed as read-only and unable to move, manage, or invest money. Banks also view AI agents as a competitive threat if they help consumers avoid fees or low-interest deposit accounts that support bank margins. Citi noted that AI agents it provides can be used without additional liability. The broader pattern—AI labs promising agent-driven services while incumbent service providers push back—will help determine which AI investments pay off.
