India generates 33% more data than the US for ChatGPT, Amitabh Kant said on Tuesday, but there’s a significant disparity in how large language models are trained globally.
“The important thing is that today we in India. I mean, if you look at OpenAI, Chat GPT, we are providing more data than the US, 33% more data than what the United States of America do,” Kant, who is a former CEO of the NITI Aayog, said during a session on the second day of the India AI Summit 2026 in New Delhi.
There’s a significant disparity in how large language models are currently trained, Kant went on to say noting that the Global South, and India in particular, is the engine room of AI development.
“These large language models are getting better and better on the basis of data from the Global South. It is essential that this contribution translates into benefits for these regions.” For AI to be truly inclusive, LLMs must move beyond English-centric models and become natively multilingual to serve diverse populations.