India’s AI future depends on fixing data, not just building models: MoSPI Secretary Saurabh Garg

As the global landscape of artificial intelligence shifts from experimentation to practical application, India's success in this domain hinges significantly on the quality of its data. Dr. Saurabh Garg, Secretary of the Ministry of Statistics and Programme Implementation (MoSPI), emphasized this point during his speech at the Nasscom Responsible Intelligence Confluence (RICON) 2026 on January 22. Garg highlighted that both government and private sectors are increasingly reliant on AI systems for crucial tasks such as resource allocation, beneficiary identification, demand forecasting, and outcome evaluation. However, the effectiveness of these systems is contingent upon the reliability and consistency of the data they utilize. "AI readiness is not fundamentally a model problem. It is a data and metadata issue," he asserted. He further elaborated that even the most sophisticated large language models and analytic engines can fail to deliver accurate outcomes if they are trained on inferior or poorly structured data. "When data is inconsistent, of low quality, lacks semantic clarity, or is trapped in formats that machines struggle to interpret, the results can be misleading," he cautioned. Garg warned that inadequate data quality can lead to significant policy missteps. For instance, a well-designed model might inadvertently exclude eligible households from receiving benefits simply due to mismatches in the dataset identifiers. "This is not a theory failure; it is a data failure," he stated. In an AI-driven landscape, merely making datasets public is insufficient. Garg pointed out that open data must be accompanied by application programming interfaces (APIs), quality standards, and clear problem definitions to be useful for AI systems. He stressed that metadata, which explains the origin, creation, and meaning of data, is becoming essential as part of the digital infrastructure. With proper metadata in place, algorithms can yield explainable results, allowing for well-informed decision-making based on the data. To enhance the AI readiness of government data, the MoSPI has been undertaking several initiatives, including the development of a national metadata structure, API design manuals, and platforms for data discovery and microdata. However, Garg acknowledged that not all data can be made openly accessible due to privacy and security concerns surrounding sensitive administrative records. He proposed the need for trustworthy infrastructure, controlled research environments, and privacy-preserving techniques that enable analytics without compromising personal information.