India must build its own AI models and strengthen data sovereignty: Experts

At BS Manthan, experts said India must build sovereign foundational AI models and boost compute, skills and data safeguards, warning that dependence on foreign systems risks digital neo-colonialism

India will need to prioritise building its own foundational artificial intelligence (AI) models and strengthen data sovereignty safeguards to reduce dependence on foreign technologies, experts said.

 

During a panel discussion at the Business Standard Manthan Summit moderated by Aashish Aryan on Wednesday, Sandeep Shukla, director of the International Institute of Information Technology, Hyderabad, Jibu Elias, country lead for India at Mozilla, and Arunima Sarkar, head of initiatives, at the Centre for Fourth Industrial Revolution (C41R) India, World Economic Forum, concurred that retaining domestic control over user and other critical data was necessary to ensure that AI uplifts and advances India’s population. 

 

“AI should evolve into a digital public infrastructure that serves as the base layer. On top of this, sector-specific data can be used to retrain models from which organisation-specific and individual-specific models can be created. In that sense, we really need a foundational model. Without the foundational model, if we are using ChatGPT, Gemini, Anthropic etc., we are giving all our data to them (foreign companies and countries),” Shukla said. 

In the race to build the latest frontier AI models, countries such as India should take the lead in Global South, according to Elias. “The country’s diversity, in terms of demography, geography, and the data, is our largest strength. So, if something works for India, it can definitely translate to other countries,” he said. 

India should also focus on AI sovereignty and competitiveness to ensure thorough diffusion of AI across the demographic of the country, Sarkar said.

 

“For AI diffusion, there are a few building blocks. We need compute, connectivity, electricity, talent and skills, and the linguistic capabilities to develop multi-lingual models, especially when we are looking at AI diffusion in the Global South,” she said, adding that countries must also forge strategic partnerships that can enable diverse, country-wide participation in the AI economy. 

 

Last week, at the India AI Impact Summit, the government announced that all major countries relevant to AI developments had signed the New Delhi Declaration. So far, nearly 90 countries and international organisations have endorsed the declaration.

 

The New Delhi Declaration on AI has placed AI diffusion and sharing of critical AI infrastructure between Global South countries at the heart of the next phase of development of the technology. Trusted AI emerged as one of the key tenets, with scientists and researchers at the Summit emphasising the need to develop trusted AI systems to enable adoption at scale.

 

“Noting that deepening our understanding of the potential security aspects remains important, we recognise the importance of security in AI systems, industry-led voluntary measures, and the adoption of technical solutions and appropriate policy frameworks that enable innovation while promoting the public interest throughout the AI’s lifecycle,” the declaration said.

 

In addition, India also became the 12th signatory to the Pax Silica initiative, an effort by the US for building reliable supply chain security of critical minerals. 

 

The experts stressed that building foundational AI capabilities would require a coordinated push across policy, academia, and industry. 

 

India’s AI ambitions, they said, cannot rest solely on application-layer innovation but must extend to chip design, high-performance computing infrastructure, and access to affordable and reliable energy to power data centres. Without these core building blocks, India risks remaining dependent on external ecosystems, they added. 

 

Data will be a primary strategic asset for any country, and failing to develop foundational AI infrastructure could lead to a new form of digital neo-colonialism, Shukla said. 

 

“If we are sending all our data, we are missing an opportunity to build our model, and they (foreign companies and countries) will now build products which are trained on our data to sell it back to us, like British colonialism. Given the current geopolitics, we have to worry about this neo-colonialism,” he said. 

 

The three panelists also noted that conversations around AI diffusion must focus on the Global South, as these countries need to catch up in terms of creating the foundational infrastructure, on top of which further models can be built.

 

“Many of the countries in the global south don't have the kind of infrastructure, especially in terms of compute, access to GPUs (graphic processing units) and all other things. For these countries, open source provides a new way of approaching. We want AI to be a digital public infrastructure. We want to position AI as like digital commons,” Elias said. 

 

Elaborating further on building indigenous AI solutions and foundational models, the panelists also cautioned that startups built merely on front-end layers using application programming interface (APIs) of large foreign models may struggle to sustain themselves, as companies such as Google, Anthropic, and OpenAI will ultimately prioritise monetising their own ecosystems. 

 

As a solution, students and entrepreneurs should focus on foundational innovation instead of chasing “easy” solutions, Shukla said. He also emphasised that India must encourage deeper technological capability rather than superficial AI integrations. 

 

Adding to the thought, Sarkar said that the country must focus on critical areas like education and health to improve the overall quality of life. 

 

While India ranks third in research and development (R&D), globally, the gap between the second and third player is huge and the country must make substantial progress in that sphere, according to Elias. 

 

“While government investment is important, the private sector should also invest in R&D and infrastructure building. They cannot simply take subsidies from the government to do R&D,” Shukla added.

 

Delving into the importance of skilling and talent development, the panelists pointed out that while India produces a large pool of engineers annually, there is a need to have specialised training. Universities, they said, must strengthen research-oriented programmes and encourage collaboration with industry to bridge the gap between academic research and commercial deployment.