How indigenous AI capabilities can strengthen India’s security and strategic autonomy

By Kaushal Bheda, Director of GovTech, Pelorus Technologies

On 12 June 2026, one of the world’s leading AI companies disabled access to its two most advanced models for every customer worldwide. The reason was a US government export control directive that cited national security and ordered that no foreign national be allowed to use them. The company could not separate foreign users from domestic ones quickly, so it switched the models off for everyone.

When a critical capability runs on foreign infrastructure and is governed by foreign law, it can be switched off overnight. India already manages this kind of risk in energy, where dependence on a single source is treated as a strategic weakness. AI now belongs in the same category, because the systems built on it will sit inside defence, governance and public services.

India should use this moment to pivot. The same AI coding tools that other countries control can be used here to build software far faster and at lower cost than before. India has one of the largest pools of engineering talent in the world, and combined with AI-assisted development, that talent can build the critical software tools the country needs in a fraction of the time it once took. The same technology that created the dependence is also the means to escape it. India can use AI to build the products that reduce its reliance on foreign software.

When India builds the software itself, it owns the code, and that ownership is the heart of the security argument. Foreign software requires trusting both what the vendor discloses and what it does not, and nation-state attackers regularly exploit known and hidden backdoors in code they did not write. When the codebase belongs to India, it can be inspected and hardened, because defenders can see what they are protecting. Owning the code turns security from an act of trust into an act of verification.

That logic does not stop at the application. A real defensive posture comes from owning the entire stack, which means the operating system, the applications and, over time, the hardware beneath them, with India’s own AI models running on top. Each layer that India owns is a layer that cannot be withdrawn or used as a point of pressure.

There is an economic gain alongside the security one. Indian IT faces pressure from tighter visa rules and from AI changing the services model that built the industry. Moving that talent toward building and owning products, rather than delivering services for others, takes the industry up the value chain. India has done versions of this before. UPI and Aadhaar showed that India can build world-class public digital infrastructure at national scale, and the same approach can extend from payments to large software systems, from the operating system layer up to the applications people use every day.

What decides whether any of this is adopted is ease of use. If a product is functional and frictionless, people will use it, and Apple is the example. The software and AI tools India builds have to be well made and simple, so that people choose them because they work better, not because they are mandated. Over-engineered, complicated products die early deaths, however sound the policy behind them.

This also means India does not have to win every part of the AI race to gain from it. The hardest layer, advanced chips and large foundational models, is already being pursued through national missions. The larger near-term opportunity is higher up, where India can build focused AI models for specific, high-value tasks in defence and security and deliver them inside products built for the Indian market. These models can be trained on our own knowledge of the threats we face, which is something no foreign vendor can supply, because we understand our own adversaries and environment better than anyone else.

This is all within India’s reach. The country has the engineering talent, and AI now gives that talent the speed to build at a scale that was not practical a few years ago. What is needed is the decision to treat software and AI as strategic infrastructure and to build it at home.