Archive · 20 Aug 2026
RBI Model Risk Rules Face Calls for Innovation Safeguards
India’s central bank has proposed rules for checking computer models used by financial companies.
These rules would apply to everything from credit-scoring systems to customer-service chatbots and spreadsheets.
The goal is to prevent models from making harmful or unfair decisions.
The authors say the rules are thorough but may be too difficult for small organisations to follow.
They also worry that strict checks could slow the use of new technology.
Supporters of strong rules say financial companies need protection because hidden problems in computer models can hurt customers.
The authors suggest making simpler rules for low-risk uses.
They also suggest testing new systems in supervised sandboxes and creating one shared testing centre.
These changes could protect people while allowing useful AI projects to continue.
The RBI’s June 24 draft would require banks, NBFCs, and cooperative lenders to govern every model they use.
The proposal draws on regulatory frameworks from the United States, Britain, Canada, and the European Union.
Critics say smaller institutions may struggle with validation, documentation, audit access, and other compliance requirements.
Supporters of a high bar point to past stress from opaque underwriting and consumers’ limited ability to challenge algorithmic decisions.
The authors propose low-risk exemptions, supervised sandboxes, and a shared national testing facility before the guidance is finalised.
- Who
- The Reserve Bank of India, banks, non-banking financial companies, cooperative lenders, and financial-sector technology users are involved.
- What
- The RBI has released draft Guidance on Regulatory Principles for Model Risk Management covering every model used by covered lenders.
- When
- The draft was released on June 24.
- Where
- The guidance concerns financial institutions in India and draws on international regulatory approaches.
- Why
- It aims to manage risks from financial models, including possible AI-related errors, bias, and opaque decision-making.
Strong Oversight
Proportionate Innovation
How high should the regulatory bar be?
Strong Oversight
A high bar is justified because India’s NBFC sector has experienced stress associated with opaque underwriting, while many retail customers cannot effectively challenge algorithmic decisions.
Proportionate Innovation
The requirements may be too costly and technically demanding, especially for smaller organisations, causing them to avoid useful technology or replace substantive oversight with paperwork.
Should all models face similar requirements?
Strong Oversight
Broad coverage can ensure that risks are not missed in less obvious tools, including pricing spreadsheets and customer-facing systems.
Proportionate Innovation
Low-risk, non-decisioning applications such as customer-service bots should receive proportionate treatment rather than the same compliance burden as high-impact models.
How should institutions build validation capacity?
Strong Oversight
Institutions should meet demanding validation, documentation, audit-access, and human-oversight expectations before deploying models.
Proportionate Innovation
Supervised sandboxes and a shared testing facility could let institutions pilot advanced models and share scarce technical expertise instead of duplicating expensive validation.
- Regulator
- Reserve Bank of India
- Draft guidance
- Guidance on Regulatory Principles for Model Risk Management
- Covered institutions
- Banks, NBFCs, and cooperative lenders
- Scope
- Every model used by covered institutions, including spreadsheets and AI models
- International sources
- Frameworks associated with the Federal Reserve, Bank of England, OSFI, and the European Union’s AI Act
- Proposed safeguards
- Low-risk exemptions, regulatory sandboxes, and a shared national testing facility