In this policy paper, Radhika Radhakrishnan and I propose a regulatory framework with thresholds for intervention that seeks to overcome the challenges in preserving transparency when dealing with machine learning algorithms, and suggest solutions to building interpretable models right from the design stage.