deniz.in

Markets

Weather

Loading weather

· via TechCrunch

Citibank puts liability for third-party AI agents on customers, even after hacks

Citibank now holds customers responsible for anything third-party AI agents do with their accounts, including losses after credential leaks, as other major US banks adopt the same stance.

Citibank puts liability for third-party AI agents on customers, even after hacks

Citibank, the third-largest commercial bank in the United States, has updated its online terms to make customers solely responsible for whatever third-party AI agents do with their accounts — including damage that follows when those agents are hacked, TechCrunch reports.

What the new terms say

The change targets the emerging class of AI tools, from general-purpose frontier models to dedicated personal finance assistants, that users hand their banking credentials so the software can analyze spending, offer investment advice, pay bills or carry out transactions.

According to TechCrunch, a Citi spokesperson said the bank revised its Online User Agreement to clarify the risks of sharing credentials with, or granting account access to, third-party AI tools, and to reinforce that customers who do so are responsible for everything those agents do.

The most consequential detail concerns security failures. Under the updated terms, if a third-party AI tool leaks a customer's credentials or is itself breached, the customer — not the bank — is on the hook for whatever attackers do with the account afterward.

Citi is not alone

The policy is the latest in a fast-moving wave. JPMorgan and Bank of America, the country's two largest banks, announced similar policies last month. PNC, the eighth largest, changed its terms on the day TechCrunch published its report. Citizens, another major US bank, emailed customers this week with the same warning, according to a message seen by TechCrunch.

Citi's terms include one notable carve-out: AI agents the bank itself provides can be used without additional liability, meaning its own tools escape the rules now applied to outside agents.

Startups face a trust gap

For AI startups hunting new revenue, personal finance looks attractive — a sector that generates steady fee streams and where an assistant could visibly add value. But consumers already hesitant to share email or chat records with an AI agent now face a far higher trust bar with bank and brokerage accounts.

Anthropic's approach is illustrative. The personal finance tool it launched earlier this month is explicitly billed as read-only and unable to move, manage or invest money — a design that sidesteps the transaction liability banks are now shifting onto customers.

Banks defending turf

Security is only part of the motivation. Banks also appear to view AI-driven finance as a competitive threat to their franchise. TechCrunch reports that markets briefly shuddered at the prospect of AI tools helping consumers sidestep unnecessary fees or low-yield deposit accounts that pad bank margins. If agents become a primary interface for financial decisions, banks intend to control that pathway — and liability terms are one lever for doing so.

The dynamic fits a broader pattern: AI labs promise agents will transform how people engage with service providers, and the providers push back. Ride-hailing and food service have already seen versions of this friction, and how such relationships resolve will help determine which AI investments actually pay off.

Why it matters

The shift moves real financial risk onto early adopters of AI agents. Anyone who grants an assistant access to a bank account now does so knowing the bank may decline to reverse agent-initiated transactions — or even fraud that follows a credential leak at the agent. That changes the calculus both for consumers and for startups building agentic finance products, which must either stay read-only, as Anthropic has done, or find ways to stand behind their agents' actions themselves. It also hands incumbent banks a structural advantage for their own agent offerings, and sets a template other industries confronting AI agents are likely to copy.

  • #ai-agents
  • #banking
  • #citibank
  • #fintech
  • #liability

Related posts