Mint Explainer: RBI proposes structural controls, use case-based approach for AI/ML models
“Kill switch mandates, independent validation of third-party models, mandatory explainability requirements, and human oversight of automated decisions are now regulatory baselines rather than best practices,” NPST’s Kumar said. Model accuracy, long considered the primary measure of AI investment, is now a necessary but insufficient condition, he said, adding that how a model is explained, audited, and overridden matters just as much as how well it performs.
Post Comment