On 6 March 2025, the Government of India unveiled AIKosha (also styled “AI Kosha” or “AIKosh”) — a sovereign datasets and AI-artefact platform — as a flagship initiative under the IndiaAI Mission. This new platform is intended to democratize access to high-quality, non-personal datasets, models, use cases, and development tools, thereby catalyzing AI research, innovation, and applications across India. The launch marks a critical step toward reducing dependence on foreign AI models and enabling homegrown, India-centric AI development.
The National AI Portal Of India
Vision & Purpose of AIKosha
One of the core challenges in AI development is a scarcity of high-quality, clean, ethically sourced, India-relevant datasets. Many existing models are trained on Western or generic datasets, which may underrepresent Indian languages, regional diversity, or local contexts. AIKosha is conceived as a centralized, curated repository of non-personal, anonymized datasets, models, toolkits, use cases, and sandbox environments aimed at bridging this gap.
By making such resources available to researchers, startups, academia, government bodies and civil society, AIKosha is expected to lower barriers to entry, reduce duplication of effort, and promote innovation tailored to India’s linguistic, geographical, social, and economic diversity.
Key Features & Architecture
Dataset & Model Repository
At its core, AIKosha houses a growing corpus of datasets and AI/ML models. At launch, over 300 datasets and 80+ models were reported to be part of the repository. These range from linguistic corpora (for Indian languages), geospatial data, governmental data, and sectoral datasets contributed by various ministries and institutions.
Contributors (academia, industry, startups) can submit Expression of Interest (EOI) to contribute their datasets, models, or use-case artefacts, along with documentation and metadata, under agreed terms. India AI
Sandbox & Integrated Development Environment
Beyond static repositories, AIKosha provides a sandbox environment — a development and experimentation space where users can build, test or refine models using the datasets and tools hosted there. It also supports discovery metadata, permissioned access, security controls (encryption in transit and at rest), APIs, and access controls to ensure safe operations.
Interlink with Compute Infrastructure
AIKosha is not standalone; it’s designed in tandem with the AI Compute Portal under the IndiaAI Mission. While AIKosha offers the data and models, the Compute Portal provides subsidized access to GPU and cloud compute resources for training and inference, thus completing the data-to-model-to-deployment pipeline.
At launch, about 14,000 GPUs were deployed via the Compute Portal, with more in the pipeline. The government also has plans to scale this further and even develop domestic GPU manufacturing over the next 3–5 years.
Role within the IndiaAI Mission & Impacts
Positioning within Mission Pillars
AIKosha corresponds to the “Datasets / AI Artefact” pillar of the IndiaAI Mission, one of several core verticals (others include Compute, Innovation Centre, Startup Financing, Application Development, FutureSkills, Safe & Trusted AI, etc.) The mission, approved by the Indian Cabinet on 7 March 2024, has an outlay of around ₹10,371.92 crore to fuel AI development across sectors.
By enabling easier access to non-personal, standardized datasets and models, AIKosha is intended to democratize AI access, especially for smaller players, and to help India reduce dependence on foreign AI solutions.
Boosting Innovation & Startups
Startups, research groups, and academic institutions can leverage AIKosha to prototype and validate AI models more rapidly. It removes the hurdle of collecting or cleaning large datasets from scratch. Also, by combining with the Compute Portal, even resource-constrained innovators can access compute resources.
To stimulate model development, the IndiaAI Mission has already received 506 proposals (as of April 2025) for building foundational AI models, and has selected Sarvam AI to lead the development of India’s indigenous Large Language Model (LLM).
Ethics, Safety & Responsible AI
AIKosha will also host responsible AI (RAI) solutions developed under the IndiaAI Mission’s “Safe & Trusted AI” pillar. In fact, four such RAI solutions are slated to go live on the AIKosha portal between September and December 2025. By centralizing artefacts under governance norms, AIKosha supports transparency, traceability, and accountability of AI systems.
Challenges, Risks & the Road Ahead
Data Quality & Diversity
A major challenge will be ensuring that datasets are representative, unbiased, well-labeled, and cover India’s multi-lingual, multi-cultural, multi-geographic diversity. Ensuring quality and consistency across contributions is nontrivial.
Privacy, Consent & Governance
Though AIKosha is restricted to non-personal and anonymized data, ensuring robust privacy, preventing re-identification, and managing data usage permissions will demand careful policy and technical guardrails.
Sustainability & Incentives for Contributors
To attract contributions (from government ministries, private sector, academia), AIKosha must provide incentives — recognition, credit, or potential downstream benefits. Otherwise, contributions may remain limited.
Scaling Compute & Infrastructure
While the initial rollout of GPU infrastructure is encouraging, demand may outstrip supply, leading to bottlenecks. Also, scaling into newer modalities (vision, multimodal, video) will stress infrastructure further.
Adoption & Ecosystem Engagement
Success depends on uptake — researchers, startups, government bodies must find real value in the platform and build applications using it. Outreach, training, developer support and community-building will be essential.
Important References:
Reference 1: Press Information Bureau (PIB) – MeitY launches AIKosha under IndiaAI Mission Link
Reference 2: Times of India – Government launches AI data & compute platforms, plans to build GPUs in 3–5 years Link
Reference 3: IndiaAI Official Portal – AIKosh Platform Overview & EOI Guidelines Link
Reference 4: Economic Times – Explained: IndiaAI Compute Portal, AIKosha and other initiatives under the IndiaAI Mission Link
Reference 5: IndiaAI – Expression of Interest (EOI) to Contribute Datasets to AIKosh Link
Reference 6: Economic Times – Four of IndiaAI’s Responsible AI solutions to go live on AIKosha from September Link
Reference 7: Vajiram & Ravi – IndiaAI Mission Overview Link
Reference 8: PIB – Building India’s Foundational AI Models under IndiaAI Innovation Initiative Link
Reference 9: FoneArena – MeitY AIKosha launched to accelerate AI innovation in India Link
Reference 10: Indianness – India AI Governance Mission: Pioneering Ethical and Inclusive AI Link
Follow our AI4Planet Weekly News page and IndiaAI Mission page for more updates.
