The Reserve Bank of India, the country’s central bank, has issued a draft framework for managing credit model risk.

It would apply to commercial banks, cooperative banks, non-banking financial companies, including housing finance companies, and all-India financial institutions.

The proposal would require board-approved policies across the full model life cycle, inventories of approved in-house and outsourced models, and approval from a board subcommittee before deployment or material changes.

Regulated entities would also need independent validation before deployment, after material changes or events, and at least once a year.

Reviews would cover assumptions, data, back-testing, documentation, and potential bias or discrimination.

Third-party models would be subject to the same standards, while regulated entities would remain accountable for outcomes.

Vendor contracts would also need to allow supervisory access to technical documentation and model reviews.

The draft does not separately mention generative AI, but its definition of credit risk models would include AI and machine-learning tools used in credit decisions.

This appears to be the RBI’s first explicit system-wide proposal on credit model risk, building on existing know-your-customer rules that encourage AI and machine learning for monitoring.

🔗 Source: Reuters

Recent Reserve Bank of India developments