No More Borrowed AI, India Needs Independent Models With "Kill Switch"

AI will soon sit at the core of how a nation governs, heals, teaches, feeds and delivers justice to its citizens. And in the coming days, India will need an indigenous, sovereign artificial intelligence architecture, since foreign controlled algorithms cannot be allowed to mediate these decisions. The future of 1.4 billion people cannot be entrusted to systems controlled elsewhere. For India, sovereign AI is not a choice driven by pride, but by necessity. India cannot gamble its governance, economy and civil liberties on artificial intelligence where the ultimate authority, the "kill switch", is not firmly in Indian hands. 

This is the argument India is making at the world's most consequential gatherings on Artificial Intelligence, the India AI Impact Summit: That a country which represents one sixth of humanity cannot depend entirely on artificial intelligence models built elsewhere.

Speaking to NDTV on the sidelines of the summit, S Krishnan, Secretary, Ministry of Electronics and Information Technology (MeitY), made it clear that India's approach to AI has evolved after intense debate.

"I think the idea which many people floated initially was that India should not attempt to build a sovereign model of its own and instead use open source models while focusing on applications," Krishnan said. 

"That was fairly sound advice. But after detailed discussions with stakeholders and experts within the country and outside, we came to the conclusion that there are some areas where India needs sovereign capacity in building models," he said.

At the heart of India's AI push is the idea of sovereignty - over data, language, culture, and ultimately strategic control. Crucially, India is not chasing scale for its own sake. 

Krishnan stressed that India does not need only giant, extremely expensive large language models. "These models need not necessarily be very large. We can focus on niche areas," he said. 

The Bias and Language Problem

One of the strongest arguments for sovereign Indian AI, according to Krishnan, is bias -- linguistic, cultural, and contextual.

"Existing AI models do not necessarily reflect all the data available in regional languages and the languages of India," he said. 

Much of India's linguistic richness, he pointed out, is not adequately represented on Internet, which means global models trained primarily in English and using western data sets inevitably fall short.

India has more than 20 official languages and hundreds of dialects, alongside deep cultural and social complexity.

"We want models that are sensitive to linguistic and cultural concerns in India," Krishnan said. 

For a country with vast diversity, translation layers alone are not enough, AI systems must understand India natively, not as an afterthought.

Who Holds the Kill Switch?

The second reason for sovereign AI, Krishnan argued, is strategic control. "As AI systems mature, you do not want to be in a situation where somebody else holds the kill switch," he warned. 

Dependence on foreign AI infrastructure could eventually translate into vulnerabilities in governance, security, and public services.

This is why India's push goes beyond models to the entire digital stack, from data centres and cloud services to applications and platforms.

"We need sovereignty in every stack, so data is fully in our control and usage is controlled by Indian entities," Krishnan said.

BharatGen: AI Built for India

Among the most significant announcements at the summit is of BharatGen, a government funded, IIT Bombay-led sovereign AI initiative. Krishnan confirmed that BharatGen is fully sovereign, funded entirely by the Government of India and implemented through a not for profit special purpose vehicle.

BharatGen has unveiled a 17 billion parameter multilingual foundation model, designed to work across 22 Indian languages, and trained extensively on India centric data. The model is aimed squarely at public interest applications, governance, healthcare, education, agriculture, and citizen services.

Asked how BharatGen compares with global AI systems such as ChatGPT, Claude or Gemini, Krishnan was cautious but confident.
"In terms of the benchmarking they have put out, they are able to get results comparable with the best in the business," he said. The focus 
is now on scaling these systems to population level.

Sarvam: Another Indian Bet

India's sovereign AI strategy is deliberately plural, not monopolistic. Under the IndiaAI Mission, the government is supporting multiple Indian efforts. One of them is Sarvam, incubated at IIT Madras, and backed by MeitY to build large language models specifically designed for Indian languages and voice based interaction.

"The idea is to support Indian ventures, Indian institutions and Indian start-ups," Krishnan said, "We don't want to exclude anything". 

Sarvam's models are designed for reasoning, multilingual fluency and population scale deployment, a critical requirement in a country where AI must serve not millions, but hundreds of millions.

India's Digital Public Infrastructure Advantage

India's sovereign AI ambition is inseparable from its Digital Public Infrastructure (DPI), Aadhaar, UPI, and digital governance platforms that already operate at unprecedented scale. This foundation drew strong international praise at the summit.

Addressing the gathering, French President Emmanuel Macron described India's digital journey as unmatched. "India built something that no other country has built, a digital identity for 1.4 billion people," Macron said, praising India's open, inter-operable and sovereign digital systems. 

He pointed to UPI and digital public platforms as proof that India has already demonstrated how technology can be deployed inclusively at scale.

Macron's message has reinforced India's argument: A country that has already built digital rails for over a billion people is uniquely positioned to shape sovereign, responsible AI.

Modi's Global Pitch

Prime Minister Narendra Modi, inaugurating the summit, placed India's AI push in a global context. "Any AI model that succeeds in India can be deployed anywhere in the world," PM Modi said, arguing that India's diversity, scale and complexity make it the ultimate testing ground for robust AI systems.

PM Modi also revealed that three Indian companies have launched AI models and applications at the summit, showcasing what he called the talent and depth of India's youth led innovation ecosystem. His message was clear: What works for India can work for the world.
India's AI vision, he stressed, is not about monopolies but about democratisation. AI, he said, must become a multiplier, not a monopoly, and must serve for inclusion rather than concentrate power.

AI for India, AI for All

Walking through the summit halls, the scale of participation underscores that vision. Nearly 1,000 start-ups, many founded by young Indians barely out of college, are showcasing AI solutions built for local problems -- from farming and healthcare to education and governance.

As India sets its sights on becoming a developed nation by 2047, the argument being made in New Delhi is straightforward: AI will shape that journey, but only if it is sovereign, inclusive and grounded in India's realities.

For a nation of 1.4 billion people, speaking dozens of languages, generating vast volumes of data every day, and already operating one of the world's largest digital ecosystems, dependence on imported intelligence is not an option.

India is no longer asking whether it should build sovereign AI models. The question now is how quickly they can be scaled, not just for India, but for the world.

India's IT minister Ashwini Vaishnava has emphasised: "We should not remain dependent on anybody for our strategic requirements."