RBI mandates kill switch for AI models at banks, introduces comprehensive model risk framework

The RBI has released draft guidelines mandating banks to implement a comprehensive Model Risk Management Framework for AI and ML systems, including kill-switch mechanisms to immediately deactivate models producing harmful outputs. The framework introduces risk-based tiering of all models, board-level accountability via a Board-approved MRM framework, and full accountability for third-party and vendor-sourced AI models. High-risk models will require approval from the Board's Risk Management Committee, and banks must address explainability, bias, fairness, and supply chain concentration risks.