Pathways to AI Sovereignty for India

It is the season of technology sovereignty, and globally, a narrative about the need for nations to reclaim control over their critical technologies and infrastructure is gaining momentum. 

The current trend of geopolitical tensions layered with technology export controls, decoupling research collaborations, and investment restrictions, among other things, started with the US seeking to convert its technology dominance into strategic leverage against a fast-rising deputy superpower in China. The denial regime indicates the US's view that technology sovereignty is a zero-sum game. Despite this, technology diffusion did happen, and in the field of artificial intelligence (AI), China's DeepSeek breakthrough is a testament to this.

What does it mean, then, for a country like India, still playing the catch-up game in capability-building, to achieve technological sovereignty in rapidly evolving general-purpose technologies such as AI, especially in a complex and interconnected world?

Multiple clarion calls have been made for India to be an AI sovereign. A theme of nationalistic fervour for self-sufficiency often underpins these. This is sure to take decades in time. Even then, no nation may be 100% self-sufficient.

AI is an enabling technology and must not be seen as an end in itself. It is critical to support our polity's economic well-being, prosperity and national security needs. As posited by Jacob Edler and others, technology sovereignty is about the ability of a State to act independently and serve the needs of its people in the face of institutional and economic boundary conditions, including third parties' adversarial actions. Christoph March and others, too, refer to technology sovereignty as "ability", not autarky or self-sufficiency.

Viewing technology in these instrumental terms may reveal some answers for India's AI sovereignty. A country may not need to be world-leading and wholly indigenous in its technology capabilities to pass as a tech sovereign. 

Pathways to AI Sovereignty for India

A clear pathway to sovereignty in AI, as in any other technology, lies in a committed innovation policy that invests in research and development (R&D) and education, promotes the development of private markets and creates a conducive ecosystem for effective technology diffusion across multiple sectors. 

Any strategy lacking such a holistic and systemic perspective would merely be a reactive, whack-a-mole approach, returning us to the drawing board to scramble together resources to fight the first fight posed by every seminal technology advancement.

However, this entails long timelines, necessitating a more nuanced take on AI sovereignty over the immediate to medium-term future. 

Building base-level capability across the value chain of computing, data, model, and application is an essential shield against the theoretical worst-case scenario in which India is wholly cut off from foreign access to all AI. Such capability would also spark in-house development and eventually compound over decades. 

The question is how much of this should be led by state intervention. 

The IndiaAI mission, with an ambitious budgetary outlay of Rs. 10,372 cr over the next five years, is still modest compared to Big Tech's estimated spend of $189 billion globally in 2024 alone. Yet, the mission is spread too thin across seven AI pillars, with 44% of the funds being earmarked for compute capacity and 19% for innovation centres focused on large multimodal and domain-specific models. While necessary, this risks being a venture capital-style spray-and-pray approach, with the confused and incoherent narrative being about building India's singular breakthrough with a DeepSeek-like moment. This needs a re-shift. In light of India's low state capacity, the IndiaAI mission should be prioritised more effectively and optimised to serve a singular goal of capability-building in the immediate term. It must seek to play an enabling role with minimum intervention by focusing on a few key pillars, such as creating a unified data platform that private players could then leverage.

The role of the State should be beyond the narrow and piecemeal focus on GPUs and large language models. The key to AI sovereignty may lie in pursuing asymmetric options and being sufficiently good, if not dominant, in one or more of the AI value chain components so as to be entrenched enough in the global supply chains for other players not to act against India's interests. India's strengths have traditionally been its sizable global footprint in the services industry and its human capital in software and AI. Unfortunately, India risks being vulnerable in both these advantages unless these strengths are tailored for an AI era with effective re-skilling and adaptations. This should be prioritised over the medium-term timeline.

Another asymmetric option for India is to promote and fund open-source AI research and projects, including global ones, rather than restricting itself to only those of Indian origin.

By undercutting the dominance of incumbent nations and private players, allowing a more decentralised and democratic platform for technology dissemination, and providing a way forward in strategic autonomy and independence, open-source technologies offer India critical leverage. 

India must combine this with astute foreign policy, engage in multilateral forums, and form global partnerships with like-minded parties such as the European Union, Japan, and Australia to ensure consistent and reliable access to AI. The recently concluded Paris AI Action Summit, co-chaired by France and India, was a step in this direction. Touted as the "third way", it called for a multilateral and platform approach to AI. As stated by the final summit statement, such a platform would incubate public-interest AI projects and help with resource sharing and technology transfer to developing nations. 

India's path to AI sovereignty involves doubling down on such measures to safeguard its interests. 

Countering National Security Concerns 

While building sovereign AI capabilities will require phased timelines, addressing national security concerns posed by foreign-origin technology and the absence of sovereign AI is non-negotiable. 

The lack of alternatives to mainstream social media and other digital apps in India is likely to affect the cognitive autonomy of our citizens, which could have implications for democratic discourse. The dominance of a few players in AI models, cloud infrastructure, and computational resources risks creating systemic vulnerabilities with negative externalities such as supply chain bottlenecks, hardware security gaps, and market distortions. 

India must strengthen its regulation, evaluation and governance capabilities to counter these. An urgent requirement is an Indigenous AI-adjacent talent base capable of developing and implementing transparent and auditable techno-legal frameworks. 

Acknowledgement : This blog is a product of collective deliberations with and contributions of the Takshashila research team.

Author