The Office of the Comptroller of the Currency published a 22-page document on November 4 to help its examiners assess artificial intelligence risks at US national banks. The framework, which creates no new prudential requirements, is based on three criteria: the potential consequences of a failure, the transfer of decisions to machines and the degree of remaining human oversight. Autonomous payment agents and real-time credit decisions are placed in the highest risk category, while banks are expected to verify user identities, limit permissions and transaction amounts, and provide an emergency stop. The guidance follows JPMorgan’s decision to block external AI agents, followed by similar restrictions from Morgan Stanley, Bank of America and Citi, as agent-driven cryptocurrency payments continue to expand rapidly.
Source: Read the original article
