RBI has issued draft guidance requiring all regulated banks, NBFCs, and financial institutions to put formal Model Risk Management (MRM) frameworks in place — covering every model they use, including AI/ML models. Public comments are open until July 24, 2026.
On June 24, 2026, RBI released 'Draft Guidance on Regulatory Principles for Model Risk Management.' This is the first comprehensive, entity-wide framework for managing model risk — covering the entire model lifecycle, including third-party models and AI/ML systems. It builds on the Aug 2024 credit-model draft and the FREE-AI committee report (Aug 2025).
A 'model' in RBI's framework is any quantitative tool used to estimate outcomes or make decisions — from credit scorecards and pricing engines to fraud-detection algorithms and AI/ML-powered underwriting systems. 'Model risk' is the risk of adverse outcomes when these tools are incorrect, misused, or poorly governed. RBI now requires regulated entities to treat this as a formal, board-owned risk category.
The draft covers every stage: model development, validation, approval, deployment, monitoring, and retirement. Critically, AI/ML models — including those from third-party vendors — are explicitly in scope. This means your bank's credit bureau score, NPA prediction model, or KYC verification tool must all be managed under this framework.
RBI expects entities to establish a Model Risk Management Policy (approved at Board level), designate clear ownership for every model, maintain a model inventory, and set up independent model validation. Senior management must have visibility into model performance and model-driven decisions.
Buying a model from a vendor does not transfer the risk. The draft is explicit: regulated entities remain responsible for the model risk of third-party models they use. Due diligence, ongoing monitoring, and validation apply equally to vendor-supplied AI tools.
Model Risk Management is now a Board-level regulatory expectation, not just an internal audit item. Expect questions in JAIIB, CAIIB, and bank officer promotion exams on MRM frameworks, model lifecycle, AI/ML governance, and the role of independent validation. The RBI FREE-AI committee's August 2025 recommendations feed directly into this guidance.
RBI first flagged model risk in credit specifically (Aug 2024 draft). The FREE-AI committee (Aug 2025) expanded the lens to AI governance broadly. Today's draft takes the final step: a single, entity-wide MRM framework covering all models across all business functions — credit, treasury, operations, fraud, and beyond.