Amitabh Kant urges Global South including India to build indigenous AI models using local data

Former Chief Executive Officer of NITI Aayog and India’s G20 Sherpa Amitabh Kant on Tuesday said the Global South, including India, should build its own AI models based on local data to transform the lives of citizens.

Speaking during a panel discussion on ‘AI for India’s Next Billion: Intergenerational Insights for Inclusive and Future-Ready Growth’ at the India AI Impact Summit 2026, he said AI must be multilingual, accessible, affordable, and accountable.

Advertisement

He further said the key challenge is whether AI can reach populations living below the poverty line, and whether it can be used to transform lives in the Global South and improve learning, health outcomes, and nutritional standards.

Taking to social media platform X to share his views, Kant said, “AI must be designed and governed as public infrastructure—accessible, affordable, multilingual, and accountable—much like India’s Digital Public Infrastructure. If AI does not work for low-income users, low-bandwidth environments, non-English speakers, women, farmers, MSMEs and frontline workers, then it is not fit for purpose. If designed as public infrastructure, governed with trust, and deployed at population scale, AI can become the most powerful inclusion tool of our generation.”

“AI cannot remain the preserve of a few large firms or elite consumers. It must function as a foundational capability embedded into public service delivery and everyday economic activity. Build open, interoperable rails, allow competition, and let scale dramatically reduce costs,” he added.

During the panel discussion, Kant also pointed to the growing contribution of data from India and other Global South countries in training large language models (LLMs). “India today provides 33 per cent more data than the United States,” he said.

He noted that LLMs are improving based on data from the Global South and warned that big tech firms could build business models using such data and later sell products at high costs. He argued that India and other developing countries should build their own AI models based on their own data to ensure equitable benefits.