The new U.S. banking model risk regulation, SR 26-2, replaces the 2011 SR 11-7 by shifting from prescriptive rules to a principles-based, proportional approach requiring banks to defend their chosen model risk frameworks. While offering flexibility, this change transfers the supervisory burden onto banks rather than reducing oversight. Notably, the updated rule excludes generative and agentic AI models pending future guidance, leaving banks accountable but without specific regulatory playbooks for AI systems. In contrast, regulators in Singapore, the UK, and Canada have integrated AI governance into their frameworks, emphasizing the need for oversight of adaptive AI. Early AI-related compliance failures, such as a $2.5 million settlement for biased AI lending practices, highlight the urgency for clear AI risk management in banking.