In News

The Ministry of Electronics and Information Technology (MeitY) has launched AI Kosha, a secure platform to catalyze Artificial Intelligence (AI) innovation by providing centralized access to high-quality datasets, models, and development tools. This initiative is part of the broader IndiaAI Mission, which has an outlay of ?10,371 crore and aims to democratize AI access and boost research and governance applications.

Key Features and Infrastructure

• Datasets & Models: AI Kosha hosts 316 datasets and over 80 AI models, covering areas such as Indian language translation, health, census, meteorology, pollution, and satellite imagery.
• AI Sandbox Environment: Offers integrated development tools, tutorials, and an IDE for training AI models.
• Security Protocols: Implements encryption, secure API-based access, real-time threat filtering, and tiered permissions for users (researchers, startups, government bodies).
• AI-readiness Scoring: Aids users in selecting relevant and usable datasets.

Compute Capacity Boost

Under the Compute Capacity pillar of the IndiaAI Mission, the government has commissioned 14,000 GPUs (up from 10,000 announced earlier in 2025) to support shared access for startups and academic institutions. This infrastructure is vital for training large AI models, particularly foundational models tailored for Indian needs.

Policy and Data Governance Background

• AI Kosha complements earlier government efforts like data.gov.in, which already hosts 12,000+ public datasets.
• A 2018 committee, led by Infosys co-founder Kris Gopalakrishnan, proposed access to non-personal data from private firms to promote innovation—a proposal that faced resistance from industry stakeholders.
• The platform promotes ethically sourced and consent-based datasets, aligning with responsible AI practices.

Challenges

• Limited Dataset Diversity: Current datasets are mostly government or research-based, limiting private-sector applicability.
• Access Barriers: Strong security protocols, while crucial, may hinder ease of access for some innovators.
• Early Stage Evolution: Wider participation from industry is essential to expand dataset variety and utility.