US Banking Regulator Issues Guidelines for Artificial Intelligence Use
Rules focus on operational risks and controlling external autonomous agents accessing customer accounts.
The Office of the Comptroller of the Currency has published its first supervisory guide for artificial intelligence use in the banking sector. The document directs examiners to prioritize potential harm caused by failures rather than focusing solely on the technical sophistication of models.
The move follows reports from major financial institutions regarding unauthorized autonomous agents operating within customer accounts. Banks such as JPMorgan, Bank of America, and Citi have restricted access from external software to prevent actions that exceed user instructions.
The regulator warns that agents operated by tech companies or third parties create significant risks even when the bank does not control the system. The guidelines suggest that examiners evaluate measures such as identity verification and transaction limits.
The framework classifies systems into four risk levels. Internal productivity tools occupy the lowest category, while fully autonomous payment agents and real-time credit systems are at the top of the danger scale.
High-risk systems may require independent testing and mandatory human approval before performing concrete actions. The goal is to allow the use of financial assistants without granting unlimited autonomy over customer funds.
Spotted an error? Write to suporte@criptoico.com.br.
