MoSPI Secretary Saurabh Garg Highlights Importance of Quality Data for AI Success in India

The effectiveness of Artificial Intelligence (AI) in India hinges more on the quality of data than on technological advancements, according to Saurabh Garg, Secretary to the Ministry of Statistics and Programme Implementation (MoSPI). Speaking at the Responsible Intelligence Confluence (RICON), Garg highlighted that even the most sophisticated AI systems could falter if they rely on poor, inconsistent, or inaccessible data formats.

Representative image

Garg emphasised that AI readiness is primarily a data and metadata challenge rather than a model issue. He noted that large language models and predictive systems struggle when faced with inconsistent data formats, low-quality signals, or data locked in PDFs and images that machines cannot easily process. This can lead to errors, such as wrongly excluding eligible families from government welfare schemes.

Data Quality and AI Trust

The Secretary stressed the importance of machine-readable, well-described, and semantically clear data for AI to be trusted. He stated that AI readiness involves fixing foundational data issues rather than merely adding AI capabilities. The credibility of AI and governance depends more on the integrity of AI infrastructure than on algorithm sophistication.

Garg pointed out that statistical and data institutions play a crucial role in the AI landscape. They are not peripheral but foundational to the AI revolution. The Ministry of Statistics has taken significant steps to ensure data reliability by updating frameworks, publishing national metadata structures, and issuing API design manuals.

Frameworks and Guidelines

The Ministry has launched discovery platforms and comprehensive cataloguing datasets across ministries, all accessible through their portals. They have also issued data harmonisation guidelines, working through statistical advisors in various industries and departments nationwide. This approach maintains data harmonisation with originating institutions while enforcing common quality assurance standards.

Garg noted that without properly prepared data for AI comprehension, systems might resort to alternative sources, potentially undermining trust in the original data. He highlighted the Ministry's obligations to make data reliable through updated frameworks and statistical quality assessment.

Future Data Consumption

The readiness of data for AI use is crucial not only from a policy perspective but also because future data consumption will increasingly rely on AI-driven elements. Garg expressed optimism that with sustained execution of guidelines and frameworks, public data would become AI-ready sooner rather than later.

He concluded by stating that if these efforts succeed, India will enable AI that is explainable, auditable, and trustworthy. The focus remains on ensuring that foundational data issues are addressed to build a credible AI infrastructure.

With inputs from PTI