Tyrian Trade

Major US Banks Shift Responsibility to Customers Using Third-Party AI Agents

Major US banks including Citi $C, Bank of America $BAC and JPMorgan Chase $JPM are tightening policies governing third-party artificial intelligence agents that access customer accounts, raising questions about liability when automated software makes transactions or exposes sensitive credentials.
Citi recently updated its online user agreement to clarify that customers who authorize third-party AI agents to access their banking accounts are generally responsible for activity conducted through those tools. The bank also warned that customers could face responsibility for transactions resulting from credentials exposed through an external AI service.
Bank of America issued similar guidance, cautioning customers about sharing online banking credentials with autonomous assistants. JPMorgan has taken a more restrictive approach to external AI systems performing certain account functions, reflecting concerns about security and accountability when financial software acts without direct human supervision.
The policies come as AI companies develop agents capable of navigating websites, reviewing financial information and carrying out tasks on behalf of users. While these systems may improve convenience, they can also create new security vulnerabilities if credentials are stored improperly, permissions are excessive or an agent performs an unintended action.
An important distinction remains between third-party tools independently authorized by a customer and AI services provided directly by a bank. Citi has indicated that its own authorized AI tools operate under different responsibility arrangements from external agents receiving customer credentials.
The banks' contractual language does not necessarily eliminate statutory consumer protections. Whether a particular transaction is legally authorized, and whether a customer or bank bears responsibility, can depend on the facts and applicable financial regulations. Contract provisions alone do not settle every possible dispute.
For consumers, the developments underscore the importance of understanding what permissions an AI agent receives and whether a bank explicitly supports the integration. Read-only tools generally create different risks from agents capable of initiating payments or changing account information.
The broader issue is how financial institutions will balance innovation with security as AI agents become more capable. Establishing clear authentication standards, transaction controls and liability rules will be essential before autonomous account management can become a mainstream banking service.

Major US Banks Shift Responsibility to Customers Using Third-Party AI Agents

Loading…