Tyrian Trade

US Banks Tighten AI Agent Access Rules and Warn Customers About Unauthorized Transfers

Major US financial institutions are tightening restrictions on third-party artificial intelligence agents, warning customers that allowing outside software to access their accounts could create liability for transactions made through those tools.
Citi has updated its customer terms to clarify that account holders who authorize outside AI agents may be responsible for activity conducted using those permissions, including mistakes or misuse involving credentials shared with an external service.
PNC has also tightened its rules for third-party access, while JPMorgan Chase and Charles Schwab have restricted autonomous tools from executing transactions without approved authorization. Bank of America has warned customers about the risks of sharing account credentials and allowing automated systems to initiate payments.
Wells Fargo has published updated online access terms scheduled to take effect on November 24. The language addresses the use of third-party AI tools and the bank's ability to restrict access, while placing responsibility on customers for certain actions performed through authorized connections.
The changes reflect growing concern about AI agents capable of navigating websites, retrieving information and initiating financial transactions. Unlike conventional software integrations, autonomous agents may interpret instructions incorrectly or perform actions that their users did not anticipate.
However, contractual language does not automatically eliminate federal consumer protections. Under US electronic funds transfer law, liability for genuinely unauthorized transfers depends on the circumstances and applicable regulations. Whether a transaction initiated by an AI agent qualifies as authorized may require examination of how access was granted and how the transfer occurred.
There is also an important distinction between approved banking integrations that provide limited access and external agents using customers' ordinary login credentials. Banks may permit data sharing through authorized connections while prohibiting outside software from directly moving funds.
The developments expose an unresolved challenge for agent-based financial services: determining who is responsible when software acts on behalf of a person but does something the person did not intend.
For now, the banking industry's approach is cautious. Broader adoption of autonomous financial agents will likely depend on clearer authentication methods, transaction approval mechanisms and liability standards that protect both financial institutions and their customers.

US Banks Tighten AI Agent Access Rules and Warn Customers About Unauthorized Transfers

Loading…