||Artificial intelligence (AI) is an emerging focus area of policy development in India. The countrys regional influence, burgeoning AI industry and ambitious governmental initiatives around AI make it an important jurisdiction to consider, regardless of where the reader of this article lives. Even as existing policy processes intend to encourage the rapid development of AI for economic growth and social good, an overarching trend persists in India, and several other jurisdictions: the limitations and risks of data-driven decisions still feature as retrospective considerations for development and deployment of AI applications. This article argues that the technical limitations of AI systems should be reckoned with at the time of developing policy, and the societal and ethical concerns that arise due to such limitations should be used to inform what policy processes aspire to achieve. It proposes a framework for such deliberation to occur, by analysing the three main stages of bringing machine learning (the most popular subset of AI techniques) to deployment–the data, model and application stage. It is written against the backdrop of Indias current AI policy landscape, and applies the proposed framework to ongoing sectoral challenges in India. With a view to influence existing policy deliberation in the country, it focuses on potential risks that arise from data-driven decisions in general, and in the Indian context in particular.