US Banks Restrict AI Agents in Customer Accounts as 2-3 Year Policy Overhaul Looms
Updated
Updated · InfoWorld · Oct 30
US Banks Restrict AI Agents in Customer Accounts as 2-3 Year Policy Overhaul Looms
3 articles · Updated · InfoWorld · Oct 30
Summary
Major US banks are detecting AI agents logging into customer accounts with user-provided credentials and are already restricting or blocking some of that access.
The clash centers on authority and liability: customers say they authorized the agent, while banks often bar credential sharing or nonhuman operation through consumer websites.
Banks face the sharpest risk because disputed or unauthorized transaction rules can force them to unwind losses even when an outside agent initiated the action.
Over the next two to three years, companies are expected to rewrite terms, authentication and dispute processes to distinguish humans, approved integrations and autonomous agents.
The likely end state is controlled delegation—read-only access, transaction limits, expiring permissions, audit trails and APIs—rather than blanket blocking of AI agents.