RBI Proposes AI Risk Framework, Mandates Kill Switch for Banks

RBI has proposed draft rules requiring banks and regulated entities to create board-approved AI model risk frameworks, with stronger oversight for all models, including vendor systems. The draft mandates a kill switch to quickly disable harmful AI outputs, emphasizes human oversight to curb automation bias, and requires explainability standards and independent validation. Customer-facing AI systems must disclose they are automated and offer access to a human representative.