The Reserve Bank of India's approach to AI governance has evolved rapidly. What started as broad technology risk guidelines has sharpened into specific expectations for financial institutions using machine learning.
The Explainability Mandate
RBI now expects that any AI-driven decision affecting a customer must be explainable in plain language. This doesn't mean every parameter must be disclosed, but the reasoning process must be reconstructible and auditable.
Model Risk Management
Banks must maintain model inventories, conduct periodic validation, and document drift detection procedures. The days of "set it and forget it" machine learning are over.
Data Localization Requirements
Customer financial data used to train AI models must remain within Indian jurisdiction. For institutions using global cloud providers, this means specific region-locking and encryption protocols.
Practical Compliance Steps
Start with a governance framework that documents every model's purpose, data sources, and limitations. Appoint a Model Risk Officer. And most importantly — test your explanations with actual customers before regulators do.