Several US banks are warning customers that using third-party AI agents to access their accounts could leave them responsible for what those agents, or hackers using compromised credentials, do. The shift turns a familiar question about AI advice into a practical one about who bears the cost when an agent can touch your money.
Watch Desk analysis
What happened
Citibank changed its Online User Agreement this week, TechCrunch reports, saying customers who grant third-party AI tools access to their accounts are responsible for activity conducted through them. The article says Citi’s terms also address credentials leaked by an AI tool or stolen in a hack.
TechCrunch reports that PNC changed its terms on the day of the article, while Citizens emailed customers a similar warning. JPMorgan and Bank of America announced comparable policies last month, according to the report. Citi told TechCrunch the update was intended to clarify risks associated with sharing banking credentials or authorising third-party agents.
Why it matters
AI agents are being pitched as helpers for spending, investing and payments. But when a bank says the customer may be responsible for an agent’s actions, the convenience comes with a less convenient question: who pays if the tool makes a mistake or its access is abused?
The report describes banks’ stated policies, not a legal ruling on whether every term would be enforceable in a particular dispute. Still, customers should not assume that giving an agent account access carries the same protections as using a bank’s own tools.
Our read
Giving an agent your banking credentials is not just another app permission. Read the bank’s current terms before connecting a third-party tool, and check whether it can only view information or can also move money. “It’s only helping” is not much comfort if the help has a payment button.
What to watch
- Whether more banks adopt or clarify terms covering third-party AI access.
- Whether banks distinguish read-only access from agents that can make transactions.
- How disputes are handled when an agent’s actions or compromised credentials are involved.
Discussion spark: If a bank lets customers connect an AI agent but says they bear the risk, should the bank still be responsible when that agent or its credentials are compromised?
Sources and evidence
- More banks says customers using AI agents are on their own – even if they get hacked | TechCrunch (18 November 2026, 17:32 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.