Data Quality Key to AI Success, Says MoSPI Secretary

New Delhi, Jan 22 (PTI) The success of Artificial Intelligence in India depends more on high-quality data than on the technology itself, Secretary to Ministry of Statistics and Programme Implementation (MoSPI), Saurabh Garg, said on Thursday.

Speaking at the Responsible Intelligence Confluence (RICON), Garg warned that even the most advanced AI will fail if the underlying data is poor, inconsistent, or "locked" in formats that machines cannot easily read.

"AI readiness is not fundamentally a model problem. It is a data and metadata issue. Large language models, analytics, predictions and forecasting systems cannot work reliably when data arrives at inconsistent formats, carries low quality signals, lacks semantic clarity, or sits locked in PDFs and images that machines may find difficult to understand or take action based on," the Secretary said.

He explained that if data is not organised properly, AI could make mistakes, such as wrongly excluding eligible families from government welfare schemes.

The readiness of data for AI use is essential not only from a policy standpoint but also because future data consumption will rely on AI-driven elements.


Without data prepared for AI comprehension, systems may turn to alternative sources, potentially eroding trust in the original data, he noted.

“The Ministry of Statistics has very important obligations for making data reliable. We have updated frameworks, published the national metadata structures and the statistical quality assessment framework, issued API design manuals and launched discovery platforms…along with comprehensive cataloguing datasets across ministries, all of which are available on our portals.

“We have also issued data harmonisation guidelines and are working on it through statistical advisors in all industries and departments around the country. Our approach preserves data harmonisation with originating institutions while insisting on common standards in terms of quality assurance,” Garg said.

The Secretary emphasised that for AI to be trusted, it must be based on data that is machine-readable, well-described, and semantically clear, and said AI readiness is less about adding AI on top and much more about fixing the foundational data.

“Statistical and data institutions are not peripheral to the AI revolution. They are foundational. The credibility of AI and governance will be determined not by the sophistication of our algorithms but by the integrity of our AI infrastructure. We have policy frameworks, standards and a knowledge of technical platforms.

“With sustained execution of guidelines and frameworks, the public data will become AI-ready sooner rather than later. If we get this right, we will enable AI that is explainable, auditable and trustworthy,” Garg said.