Executive summary

The Reserve Bank of India (RBI) has released draft guidance on regulatory principles for model risk management (MRM), introducing one of the most comprehensive enterprise-wide model governance frameworks issued by a banking regulator. The proposed principles significantly broaden supervisory expectations beyond traditional quantitative and credit risk models to encompass statistical models, machine learning (ML), artificial intelligence (AI) and generative AI (GenAI) deployed across banking operations.

The guidance establishes robust governance across the entire model lifecycle, including model development, validation, approval, implementation and ongoing monitoring. It also introduces explicit expectations regarding AI explainability, human oversight and contingency controls, reflecting the growing importance of responsible AI within prudential supervision.

As financial institutions increasingly rely on AI-driven decision-making across lending, fraud detection, treasury, regulatory reporting and customer service, the RBI is positioning MRM as a strategic enterprise capability, rather than a specialist validation function.

Key takeaways

Context

The RBI has proposed comprehensive principles to strengthen governance and oversight of all models used by regulated entities, reflecting the rapid adoption of AI, machine learning and advanced analytics across financial services.

Challenge

Many institutions continue to operate fragmented model governance frameworks focused primarily on credit risk or regulatory capital models. The growing use of AI and third-party analytical models introduces new risks relating to explainability, governance, accountability and operational resilience.

Implications

Financial institutions should establish enterprise-wide model inventories, strengthen independent validation capabilities, enhance governance of third-party AI models and integrate model risk management into broader enterprise risk management and operational resilience programs.

Strategic outlook

Chartis expects similar supervisory principles to emerge across Asia-Pacific and other global jurisdictions over the coming years. As AI adoption accelerates, integrated model risk management platforms will become a strategic investment priority supporting governance, compliance and responsible innovation.

Overview and context

The RBI’s proposed MRM framework marks a significant evolution in prudential supervision, extending governance beyond traditional quantitative and credit risk models to encompass all models supporting business decisions, including AI and generative AI.

The principles apply to internally developed and third-party models, requiring enterprise-wide model inventories, clear governance across the three lines of defense, independent validation and continuous lifecycle monitoring.

A key differentiator is the RBI’s explicit focus on AI governance. It expects institutions to address explainability, bias, fairness, model drift, cybersecurity and human oversight, while maintaining the ability to suspend high-risk AI models where necessary.

The framework also reinforces a fundamental regulatory principle: while models can be outsourced, accountability cannot. Regulated entities remain fully responsible for the governance, validation and outcomes of third-party models.

Chartis Insight

Chartis believes the RBI’s proposed MRM guidance represents one of the most significant global developments in enterprise model governance and one of the first comprehensive AI governance frameworks introduced by a major Asian banking regulator.

The proposal marks a fundamental shift in prudential supervision, positioning MRM as a strategic enterprise capability, rather than a specialist validation function. By bringing traditional quantitative models, machine learning and generative AI under a single governance framework, the RBI is establishing model risk management as the foundation for responsible AI adoption.

While broadly aligned with the direction taken by US and European regulators, the RBI goes further by introducing explicit controls for AI and generative AI, including explainability, bias monitoring, human oversight and model kill switch capabilities.

For financial institutions, the framework will accelerate demand for integrated MRM platforms supporting model inventories, governance, validation, ModelOps and AI lifecycle management. Chartis expects the RBI’s principles to influence supervisory approaches across Asia-Pacific and other emerging markets, making enterprise-wide MRM a strategic investment priority.

Implications and recommendations

For financial institutions

Develop enterprise-wide MRM frameworks covering governance, model inventories, validation, monitoring and lifecycle management across all quantitative, AI and machine learning models.

For Chief Risk Officers and model risk teams

Expand validation capabilities beyond traditional statistical models to include AI explainability, bias testing, model drift detection, continuous monitoring and governance of generative AI applications.

For technology vendors

Enhance model risk management solutions by integrating AI governance, ModelOps, workflow automation, validation, explainability, model inventory management and third-party model oversight within unified enterprise platforms.

For regulators

Continue developing principles-based supervisory frameworks that integrate model risk management, AI governance, operational resilience and third-party risk management into a coherent enterprise governance model.

For executive management

Treat model risk management as a strategic business capability supporting trusted AI adoption, operational resilience and enterprise-wide governance, rather than as simply a regulatory compliance requirement.