Public Procurement of AI

AI is becoming a core part of public sector transformation. Around 67% of OECD countries are already using AI to improve government functions, while AI integration is projected to help governments cut budget costs by as much as 35% over the next decade.

India is also moving towards greater public-sector adoption of AI, with procurement already emerging at both the central and state levels. In June 2026, the Ministry of Electronics and Information Technology (MeitY) released an RFP to empanel up to 20 firms to support AI-enabled modernisation of legacy government IT systems. State governments, including Rajasthan, Maharashtra and Odisha, have also initiated AI procurement.

Yet India currently lacks a centralised framework for procuring AI services. As central and state government departments look to scale procurement, this gap risks inconsistent approaches to vendor evaluation, risk management and governance, while creating uncertainty for technology providers. AI procurement requires public authorities to assess considerations including cybersecurity, human oversight, responsible AI safeguards, accountability and post-deployment monitoring.

A centralised approach can provide a common baseline while allowing procurement requirements to remain proportionate to the risk, scale and use case of different AI deployments. MeitY’s cloud procurement guidelines provide a useful precedent, establishing common standards for vendor selection, security requirements, service-level expectations and procurement preparedness. A similar approach for AI could help government departments procure more consistently, while enabling responsible adoption and greater participation from AI providers.

TQH’s research examines how governments around the world are approaching this challenge and what India can learn from their experience. The Public Procurement of AI working paper undertakes a comparative assessment of procurement approaches in Canada, the European Union, the United Kingdom, the United States and Singapore, drawing out the principles and mechanisms that could be adapted to the Indian context. It translates these international lessons into a practical framework for India – identifying how the government can create greater consistency and confidence in AI procurement while retaining flexibility for different risks, use cases and stages of deployment.

The recommendations focus on three areas: (i) identifying the right approach to AI procurement, including whitelisting and standardised evaluation metrics; (ii) leveraging existing procurement mechanisms; and (iii) developing proportionate and flexible procurement pathways. We also propose a risk-based approach that allows government departments to distinguish between pilots and scaled deployments, as well as lower- and higher-risk use cases. Such proportionality can help enable adoption and industry participation without defaulting to a one-size-fits-all approach.

The working paper ultimately seeks to support the development of standardised, central guidance for public procurement of AI in India, providing government departments with a common framework while retaining sufficient flexibility to account for differences in risk, scale and use case. 

As a working paper, the analysis may continue to evolve as AI procurement practices develop.