Policy Update
Ashi Verma
Background
India generates one of the largest volumes of data globally, yet this data remains fragmented across ministries, institutions, and private silos. This fragmentation posed a critical challenge for students, researchers, and innovators seeking authentic, machine-readable, and India-specific datasets. Consequently, reliance on Western datasets like ImageNet has been noted as a contributing factor to the limited representation of India’s linguistic diversity, socio-economic realities, and local contexts in AI models.
To bridge this critical gap, the Ministry of Electronics and Information Technology (MeitY) launched the IndiaAI Mission in March 2024, with a budget outlay of Rs. 10,371.92 crore over five years Source: PIB Press Release, 7 March 2024.
On March 6, 2025, marking the first anniversary of the IndiaAI Mission, AIKosha: IndiaAI Datasets Platform was officially launched. Derived from the Sanskrit word meaning ‘treasure’, AIKosha is envisioned as a unified, secure, and open platform for non-personal and anonymized datasets. Unlike a conventional data portal, it functions as a knowledge repository by integrating datasets with models, APIs, and sandbox tools, thereby democratizing AI development for students, researchers, and startups to build solutions for Bharat.
Functioning
AIKosh is designed not just as a data repository, but as an end-to-end ecosystem for AI development. Its functioning can be understood through four interconnected layers
(a) Unified Data Aggregation Layer : This is the core foundation. The platform collates non-personal and anonymized datasets from various government ministries, public agencies, and voluntary private contributors. At present, it covers more than 20 sectors including agriculture, health, education, smart cities, satellite imagery, weather, census, and most importantly, Indian languages through Bhashini. The platform is designed to facilitate a process of anonymization, quality validation, and AI-Readiness scoring to make datasets machine-readable and suitable for model training.
(b) Tools and Model Access Layer : AIKosha is designed around privacy and controlled data access. The platform focuses on non-personal and anonymized datasets, while access mechanisms may vary according to the nature and permissions associated with individual datasets. It operates through a secure API gateway model and promotes a voluntary data contribution model where startups and researchers can contribute their own curated datasets.
(c) Security and Governance Framework : The platform operates on a privacy-first principle. All datasets are non-personal and fully encrypted. Access is permission-based and follows a secure API gateway model. The framework ensures data contributors retain control while allowing open access for innovation. It also promotes a voluntary data contribution model where startups and researchers can contribute their own curated datasets.
(d) Community and Innovation Layer : The final layer is focused on users, especially students and developers. The platform is integrated with the IndiaAI Compute Portal to provide access to subsidized GPU resources for model training. It is designed to foster a collaborative community through initiatives such as hackathons, innovation challenges, and learning resources, where users can learn, build, and publish their contributions.
Performance
Since its official launch on March 6, 2025, AIKosh has progressed from a policy idea to a functional platform.
- Dataset Growth and Sectoral Coverage : At the time of launch in March 2025, the platform started with over 300 curated datasets. Within a few months, it has scaled to more than 1,000 live datasets and is on track to achieve a target of over 5,000 datasets. The coverage has expanded to more than 20 critical sectors including agriculture, health, climate, education, smart cities, transport, and Indian languages. Integration with Bhashini has particularly strengthened its multilingual data repository.
- User Adoption and Ecosystem Engagement : As of late 2025, the platform has recorded over 265,000+ visits with 6,000+ registered users and 13,000+ resource downloads, demonstrating early adoption. More recent reports indicate growth to over 5,500 datasets and 251 AI models, with more than 385,000 visits and 11,000 registered users (TechTimes, Dec 2025) . It has become a primary data source for IndiaAI-organized hackathons, BharatGen, and other national-level AI challenges, with adoption among academic institutions like IITs, IIITs, and NITs for final-year projects and research.
- Integration with Compute Infrastructure : A major performance indicator is its seamless linkage with the IndiaAI Compute Portal. While AIKosh provides the data, the Compute Portal provides access to over 18,000 GPUs at subsidized rates. This data-plus-compute model has significantly reduced the entry barrier for students who earlier had data but no computing power to train models.
The government has invited startups, private firms and research institutions to contribute anonymized and non-personal datasets through an Expression of Interest under a secure, permission-based model. This collaborative approach is intended to improve both the quantity and diversity of data available. The platform nevertheless faces implementation gaps, particularly around dataset standardization, documentation and user support.
Impact
The launch of AIKosh has generated a significant impact across multiple stakeholders, particularly for the youth and innovation ecosystem
Impact on Students and Hackers : This is the most direct impact. AIKosha has democratized access to AI resources. Access to high-quality, domain-specific and machine-readable datasets has often been uneven, particularly for students and smaller research or innovation teams. Now, students from Tier-2 and Tier-3 cities can download authentic, India-specific data for their final-year projects and hackathons. It has shifted the focus from data collection to actual model building.
Impact on Startups and Entrepreneurs : For early-stage startups, data acquisition is often expensive and time-consuming. AIKosha is designed to provide ready-to-use, anonymized datasets, APIs, and sandbox tools, which can potentially lower their development cost and time-to-market. It is intended to enable them to build solutions tailored for Indian markets, such as agri-tech and health-tech applications.
Impact on Indian AI Ecosystem and Atmanirbharta : The platform aims to reduce over-dependence on Western datasets like ImageNet and COCO. By enabling training on Indian data, it seeks to help create AI models that are more culturally and contextually relevant for India. However, the use of Indian datasets does not automatically guarantee higher accuracy or removal of bias, as these datasets may themselves contain inherent biases that require careful curation and validation. This effort aligns with and strengthens the vision of Atmanirbhar Bharat and Digital India.
Impact on Academic Research : Researchers now have access to validated, machine-readable, and curated datasets across 20+ sectors. This improves the quality and relevance of AI research in India and encourages interdisciplinary studies on local problems like climate change, public health, and urban governance.
Impact on Inclusive and Responsible AI : By including datasets of Indian languages through Bhashini and covering diverse socio-economic realities, AIKosh promotes inclusive AI. It ensures that AI solutions are built for all sections of Bharat, not just for English-speaking urban users, thus making AI more ethical and responsible.
Emerging Issues
While AIKosh is a transformative initiative, certain challenges need to be addressed for its long-term success
- Issue of Data Quality and Standardization : While quantity is increasing, quality remains a concern. Datasets from different ministries are often not fully AI-ready due to missing labels, non-uniform formats, and incomplete metadata. AIKosha addresses this through an AI-Readiness Score, a metadata-level indicator evaluated before publishing, based on metadata completeness, standardized formatting, data quality, and accessibility. However, its rollout across legacy datasets is still inconsistent.
- Privacy, Security and Anonymization Risks : The platform is intended to host only non-personal, anonymized data, but ensuring complete de-identification at scale is challenging. Under the current EOI, the responsibility for anonymization lies with the contributor, who must ensure DPDPA compliance. The platform does not currently document an automated differential privacy pipeline at ingestion; it relies on contributor-side anonymization with safeguards like RBAC and encryption. Any gap at the contributor end could lead to privacy risks.
- Low Awareness and Digital Divide : A major issue is the awareness gap. Most students and developers in Tier-2 and Tier-3 cities are still unaware of the existence and functioning of AIKosh. Furthermore, lack of high-speed internet, limited knowledge of APIs/SDKs, and inadequate documentation in regional languages restricts its usage.
- Sustainability of Data Contribution Model : The platform relies on voluntary data sharing by private firms. Without clear incentives, regular contribution of high-quality datasets may be limited. Long-term viability could be improved through incentives like compute credits for GPU access, tier-based access to premium datasets, and visibility/support under IndiaAI initiatives.
- Lack of Curation and Regular Updation : Many datasets are static. For sectors like weather and transport, real-time data is critical. Instead of manual re-uploads, automated API pipelines and dynamic sync with source departments, supported by domain-specific curation teams, should be implemented.
Way Forward
To make AIKosh truly effective for students and hackers, a multi-pronged strategy is required. First, awareness and accessibility must be improved. The platform should be integrated with academic curricula like SWAYAM and NPTEL, and regular campus-level workshops and national hackathons should be organized. An AIKosh Lite version with mobile compatibility and documentation in Hindi and regional languages will help bridge the digital divide for Tier-2 and Tier-3 city students. Simultaneously, strong data curation teams comprising domain experts should be established to ensure datasets are regularly updated, properly labeled, and bias-free.
Second, the sustainability of the ecosystem needs to be strengthened. A clear Contributor Incentive Model should be introduced where data contributors are rewarded with compute credits, GPU hours, or official recognition. The platform must be tightly linked with IndiaAI FutureSkills and IndiaAI Compute Portal to provide end-to-end support from learning and data access to model training. This will ensure that AIKosh evolves from a data portal into a complete innovation hub, enabling students to build responsible and inclusive AI solutions for Bharat.
References
1. Press Information Bureau. (2025, March 6). MeitY launches AIKosha: IndiaAI Datasets Platform and AI Compute Portal to advance AI innovation. Ministry of Electronics and Information Technology. https://pib.gov.in/PressReleasePage.aspx?PRID=2108523
2. IndiaAI. (2025). AIKosh: IndiaAI Datasets Platform Official Portal. Ministry of Electronics and Information Technology. Retrieved November 28 , 2025.
https://aikosh.indiaai.gov.in/home
3. Ministry of Electronics and Information Technology. (2024, March 7). Cabinet approves IndiaAI Mission with a budget outlay of Rs. 10,371.92 crore. Press Information Bureau. https://pib.gov.in/PressReleasePage.aspx?PRID=2012356
4. The Economic Times. (2025, March 8). Government invites private firms to share anonymous data for AI Kosh. https://economictimes.indiatimes.com/tech/artificial-intelligence/government-invites-private-firms-to-share-anonymous-data-for-ai-kosh/articleshow/118789234.cms
5. The Indian Express. (2025, May 31). Key step in democratising AI: IIT-B releases 16 datasets on AIKOSH. https://indianexpress.com/article/cities/mumbai/iit-b-releases-16-datasets-on-aikosh-10041787/
6. Gadgets360. (2025, March 6). Indian government launches AI Compute Portal, dataset repository AIKosha to advance AI innovation. https://www.gadgets360.com/artificial-intelligence/news/india-government-ai-compute-portal-aikosha-datasets-platform-launch-7878789
7. IndiaAI Mission. (2025, March 25). IndiaAI Datasets Platform: Expression of Interest (EOI) for AIKosh Partners for Dataset Contributions. Ministry of Electronics and Information Technology. https://indiaai.gov.in/article/indiaai-datasets-platform-eoi-for-aikosh-partners-for-dataset-contributions
8. Ministry of Electronics and Information Technology. (2025, March 6). AIKosha Features: AI readiness scoring, content discoverability and security. Press Information Bureau. https://pib.gov.in/PressReleasePage.aspx?PRID=2108523
9. Ministry of Law and Justice. (2023, August 11). Digital Personal Data Protection Act, 2023. Government of India. https://www.meity.gov.in/static/uploads/2024/06/2bf5f02e-7c07-450a-93d3-02f8a3862a6e-Digital-Personal-Data-Protection-Act-2023.pdf
About the Contributor
Ashi is pursuing B.A Program (History and Political Science) from Maitreyi College (NCWEB) South Campus Delhi University with a major in Political Science. She is passionate about governance, public policy, social issues and digital innovation, with a keen interest in research, academic writing, and data documentation. Through research and policy analysis, she aims to contribute to evidence-based policymaking and create meaningful social impact.
Acknowledgement
I would like to express my sincere gratitude to IMPRI (Impact and Policy Research Institute) for giving me the opportunity to work as an intern and write on such an important topic. I would also like to acknowledge the official MeitY and IndiaAI Mission for making authentic data available on AIKosh portal helped me to complete this article.
Disclaimer
The views and opinions expressed in this article are those of the author and do not necessarily reflect the official policy or position of IMPRI (Impact and Policy Research Institute) or any other affiliated organisation. The information presented in this article is based on publicly available government sources and is intended solely for academic and research purposes
Name of the Reviewer’s:
Pragya and Gowri Kodali
Read more at IMPRI:
Digitalising Retirement Security: Assessing the Role of eNPS in Expanding Pension Access Since 2015
Education Reform Stalled: Promises, Protests and Political Delay
