India Plans Global AI Repository and AI Kosh for Social Good

India has reiterated its plan to establish a Global Repository of Artificial Intelligence (AI) Applications and an AI Kosh dataset platform focused on social-good deployment, underscoring the government’s intention to position the country as a key provider of public digital infrastructure in the AI domain. The initiative is anchored in the Ministry of Electronics and Information Technology’s (MeitY) IndiaAI framework and is framed as an extension of India’s earlier digital public infrastructure efforts in identity, payments and data sharing.

Reaffirmation of plan in recent policy discussions

Officials from MeitY and the IndiaAI programme have, in recent months, reiterated that work is progressing on a two-track approach. The first track is the proposal for a Global Repository of AI Applications that can be deployed for public service delivery and governance use cases across jurisdictions. The second track is AI Kosh, conceived as a national-scale repository of datasets, models and related resources for AI development with a particular focus on applications that serve public and developmental objectives.

According to policy explanations shared in government briefings and technical concept documents, the repository is being designed as a structured catalogue of AI tools, models and reference implementations that can be used or adapted by government departments, public agencies, start-ups and research groups. AI Kosh is conceived as a complementary data and model layer, consolidating curated datasets, annotated corpora, and domain-specific AI assets that can be used to build or fine-tune applications for governance and social impact.

The renewed emphasis has come in the backdrop of India’s broader IndiaAI Mission, which has proposed a multi-layered architecture for AI compute, datasets, innovation and skilling. Policy statements indicate that the Global Repository and AI Kosh are intended to be core pillars of the “data and application” layer within this architecture, interacting closely with compute infrastructure and innovation programmes.

Concept of a Global Repository of AI Applications

The proposed Global Repository is being framed as a shared pool of AI solutions that address common governance challenges such as beneficiary targeting, document processing, grievance redressal, education support, health triage, agriculture extension and urban management. It is expected to catalogue ready-to-use or easily adaptable AI applications that can be deployed in different parts of India, and in principle, by partner countries that adopt compatible digital public infrastructure frameworks.

Government concept notes indicate that such a repository would typically include application descriptions, reference architectures, open-source components where available, deployment guidelines, and evaluation metrics. The approach follows the broader Indian pattern of creating reusable digital building blocks that can be combined and repurposed by different departments rather than building isolated, one-off solutions.

Officials have also indicated that the repository will emphasise transparency and documentation. This would include detailing the purpose of each AI system, the data domains used for training or operation, indicative risk profiles and recommended safeguards. The objective is to provide implementing agencies with both technical artefacts and governance guidance, so that deployments can be aligned with sectoral regulations, data protection requirements and ethical AI principles.

AI Kosh as a dataset and model backbone

AI Kosh is envisaged as a national resource for datasets, models and tools that are specifically curated for AI projects serving the public interest. Conceptual descriptions place AI Kosh as a “data commons” for AI, emphasising:

• Curated datasets drawn from public sector records, administrative databases, surveys and other government-supported data sources, with appropriate anonymisation and access controls.
• Domain-specific datasets in priority areas such as languages, healthcare, agriculture, education, urban services, logistics and climate-related information.
• Baseline AI models, including language models and domain models, that can be fine-tuned for specific governance or social-good applications.
• Tools and documentation for data annotation, quality assessment and responsible data use.

The initiative builds on earlier Indian efforts around open government data and language technology, including work under the National Language Translation Mission, language corpora for Indian languages and sectoral data platforms in health, agriculture and urban governance. AI Kosh is positioned as a way to bring these threads together into an AI-ready resource layer, with consistent standards and interfaces.

In government briefings, officials have highlighted that AI Kosh is expected to support multilingual AI solutions that work across India’s linguistic diversity, as well as models that can work with noisy, incomplete or heterogeneous datasets that are typical of many public sector systems. The emphasis on inclusive and low-resource contexts is presented as a distinguishing feature of India’s approach to AI for social good.

Implementation architecture and institutional responsibility

The implementation of the Global Repository of AI Applications and AI Kosh is expected to be led by MeitY in coordination with the IndiaAI programme and its implementing agencies. Technical execution will likely involve a combination of government technology organisations and public sector units, along with private and academic partners selected through competitive processes.

Conceptual designs described by officials point to a layered architecture. At the base is secure cloud and compute infrastructure, which may leverage the compute capabilities proposed under the IndiaAI Mission. On top of this infrastructure, AI Kosh would provide secure storage, discovery tools and access management for datasets and models. The Global Repository of AI Applications would sit as an application and solutions layer, offering catalogues, dashboards and integration interfaces for agencies seeking to deploy AI applications.

Governance structures are expected to include expert committees or working groups responsible for vetting datasets, accrediting AI applications for inclusion in the repository, and periodically reviewing standards. These structures are intended to ensure that the repository and AI Kosh evolve with technological developments and regulatory requirements.

Government statements and positioning

In its public communication on AI, the Government of India has consistently linked AI initiatives to the broader agenda of digital public infrastructure and social inclusion. Senior officials have framed the Global Repository and AI Kosh as instruments to reduce duplication, share best practices and accelerate safe and effective deployment of AI in public service delivery.

India’s digital public infrastructure journey has demonstrated the value of reusable platforms in payments, identity and data empowerment. Our AI initiatives build on the same principle of shared building blocks, so that innovations for social good can be scaled rapidly and responsibly.

Public briefings have also underlined that these initiatives are not only targeted at central government ministries but are designed to be used across state governments, local bodies and public institutions. The intent is to create a common resource that smaller departments and states with limited technical capacity can draw upon, instead of each entity having to build AI capabilities from scratch.

Use cases in social-good deployment

The reiteration of the plan has been accompanied by references to a range of use cases that the repository and AI Kosh are expected to support. In social protection and welfare delivery, AI tools can assist in data deduplication, eligibility verification, address standardisation and targeted outreach, potentially improving the accuracy and timeliness of benefits.

In healthcare, AI-driven triage tools, diagnostic support for frontline health workers, appointment scheduling optimisation and supply-chain analytics for medicines and vaccines are among the applications that have been discussed in policy documents and pilot projects. AI Kosh’s health datasets and models would serve as a foundation for such applications, subject to strict privacy and regulatory safeguards.

Education is another key domain, where AI applications such as personalised learning support, automatic content translation into Indian languages, assistive tools for children with disabilities and learning analytics dashboards for teachers are being explored. The repository is expected to include reference solutions and toolkits that education departments can adapt to local curricula and languages.

For agriculture and rural development, AI-enabled crop advisory systems, pest and disease detection tools, weather-linked advisory services, and market information platforms are seen as potential beneficiaries of AI Kosh’s datasets in weather, soil, cropping patterns and market prices. The Global Repository could offer model applications that state agriculture departments and extension agencies can deploy with limited customisation.

Data governance, privacy and ethical safeguards

The design of AI Kosh and the Global Repository is being discussed alongside emerging frameworks for data protection, AI ethics and algorithmic accountability. Government communications have acknowledged that large-scale aggregation and sharing of datasets and AI models must be accompanied by strong privacy and security safeguards.

Policy documents suggest that datasets in AI Kosh will be subject to anonymisation, differential access controls and purpose limitation principles. Sensitive personal data will either not be included or will be made available only under strict conditions for specific research or policy purposes, in line with applicable data protection laws and sectoral regulations.

For the Global Repository of AI Applications, officials have spoken of classification of applications based on risk categories, with higher scrutiny and mandatory safeguards for systems that have significant impact on individuals’ rights or entitlements. Explainability, auditability and human-in-the-loop mechanisms are among the generic safeguards being emphasised for many classes of public sector AI deployments.

There is also an expressed focus on systemic bias and fairness. Given the diversity of Indian contexts, AI Kosh’s datasets and models are expected to be assessed for representativeness across regions, languages, socio-economic groups and genders. The repository’s inclusion criteria for applications are likely to require documentation of limitations, tested contexts and known biases.

International collaboration and “global” orientation

India has repeatedly highlighted the “global” aspect of the proposed repository in international forums on digital public infrastructure and AI governance. The concept aligns with India’s broader effort to share its public digital platforms with partner countries through technical collaborations, capacity-building and open standards.

In this context, the Global Repository of AI Applications is being positioned as a resource that other countries, especially those with similar development priorities, can access or emulate. The idea is that solutions developed for Indian conditions, including low-bandwidth settings, multilingual populations and heterogeneous data systems, may be relevant for a wide range of partner countries.

International cooperation could also extend to contributions to AI Kosh. Academic institutions, multilateral organisations and partner countries may, subject to governance frameworks, contribute datasets, models or tools that fit the social-good objectives and technical standards of the platform. Such collaboration would be contingent on compatible legal and ethical frameworks.

Potential administrative impact within India

From an administrative perspective, the Global Repository and AI Kosh are intended to change how public agencies plan and implement AI projects. Instead of commissioning isolated solutions on a case-by-case basis, ministries and departments can consult a catalogue of vetted applications, reference implementations and datasets.

This can potentially reduce procurement and development timelines, as agencies adopt or adapt existing components rather than starting from a blank slate. It may also support better interoperability between AI systems used by different departments, if common standards, APIs and data formats are enforced through the repository’s inclusion criteria.

The availability of AI Kosh as a common data and model backbone may also improve coordination between central and state-level digital initiatives. States could leverage centrally curated datasets and models while adding their own local datasets where needed, thereby creating a federated but harmonised AI data ecosystem.

Capacity-building is another area where the repository could have administrative impact. Training programmes for civil servants, public sector technologists and implementing partners can use the repository and AI Kosh as practical teaching tools. Officials can be exposed to concrete examples of AI use in governance, along with documentation of risks and safeguards.

Implications for start-ups, research institutions and industry

The repository and AI Kosh are also expected to influence the innovation ecosystem around AI for public good. Start-ups and small technology firms working on governance and social-impact solutions may benefit from access to curated datasets and baseline models, reducing entry barriers and development costs.

Public tenders and innovation challenges could reference the repository and AI Kosh as common resources, allowing innovators to focus on value-added layers such as domain adaptation, user interface design and integration with legacy systems. Over time, successful solutions from start-ups and research groups could themselves enter the repository, creating a feedback loop between innovation and deployment.

For academic and research institutions, AI Kosh can provide a structured source of real-world datasets relevant to policy and development questions. This may stimulate applied research on topics such as fairness in welfare targeting, optimisation of public transport, disease surveillance, learning outcomes and climate resilience.

Industry bodies and larger technology firms may participate through standard-setting, open-source contributions, and joint pilots with government agencies. Their involvement would be subject to clear governance rules on data sharing, intellectual property, and use of public datasets for commercial purposes.

Integration with digital public infrastructure and future roadmap

The Global Repository of AI Applications and AI Kosh are being framed as integral components of India’s broader digital public infrastructure stack. Integration points are likely with identity systems, digital payment platforms, consent-based data-sharing frameworks and sectoral registries.

Over time, as new DPI layers emerge in sectors such as health, agriculture and logistics, AI applications built on top of these platforms could be documented and shared through the repository. AI Kosh would in turn evolve as these sectors generate new datasets, signals and feedback loops that can be used to improve models.

Officials have indicated that the roadmap for these initiatives will be iterative. Initial phases may focus on a limited set of high-priority domains, with a curated set of datasets and applications. As governance structures, access protocols and technical standards stabilise, the coverage of AI Kosh and the repository is expected to expand.

Monitoring and evaluation mechanisms are likely to be built into the roadmap, tracking not only technical metrics but also the impact of AI deployments on service delivery, inclusion and administrative efficiency. Findings from these evaluations can feed back into the design of the repository, the curation of AI Kosh and the guidance provided to implementing agencies.

Challenges and areas for further development

While the reiteration of the plan signals policy continuity, several implementation challenges remain. These include ensuring high-quality, well-documented datasets at scale; harmonising data standards across departments and states; and building sustained institutional capacity for data stewardship and AI governance.

Interoperability between legacy systems and new AI-driven tools is another area requiring sustained technical effort. Many public sector databases and workflows were not designed with AI integration in mind, and retrofitting them may require both technical and organisational change.

The success of the Global Repository and AI Kosh will also depend on robust mechanisms for feedback and grievance redressal from citizens and frontline functionaries. As AI tools begin to influence eligibility decisions, prioritisation of services or allocation of resources, mechanisms to detect and correct errors or biases will be essential.

Despite these challenges, the reiteration of India’s plan for a Global Repository of AI Applications and AI Kosh indicates that the government sees shared AI infrastructure as a central plank of its digital governance strategy. The coming years will determine how these concepts translate into operational platforms, and how effectively they support the objective of using AI for social-good deployment at scale.