Why Reliance Jio AI chief says India must stop building on borrowed AI intelligence

India isn’t innovating in AI; rather, it is renting intelligence to build upon. That was the underlying theme of the thought-provoking insights shared by Gaurav Aggarwal, chief AI scientist at Reliance Jio.

India is currently focused on building applications on top of LLMs (large language models) that have already been developed in the Global North. This, Aggarwal argued, needs to change. He was speaking at the annual conference ‘State of India’s Digital Economy’ organised by think-tank ICRIER and tech investment firm PROSUS.

Read Full Story

“We can’t afford to rent intelligence. AI is different from software. Software is a tool that you buy once and use forever. AI, however, is like a raw commodity. It learns and becomes smarter over time. We are currently building our national digital fabric on foreign APIs (application programming interfaces). It is almost like leasing a tool where the landlord dictates the terms and we have to keep paying for it,” Aggarwal said. He went on to describe this as a “sovereignty crisis in the making”.

“Intelligence today is highly securitised. These are heavily guarded corporate and geopolitical assets. This capability can be curtailed if it is built on foreign platforms whose servers are located in the Global North,” he said.

To put this into perspective, Aggarwal pointed out that when the Ukraine war began in 2022, companies such as Google Pay and McDonald’s suspended their operations in Russia. “This is how geopolitics can shape corporate strategy and derail a country’s plans if it is dependent on foreign capital,” he said. “We can’t be truly sovereign if the code on which we are building applications is owned and controlled by other nations.”

Prime Minister Narendra Modi recently urged citizens to reduce their consumption of cooking oil, curb gold purchases and avoid unnecessary foreign leisure travel to preserve foreign exchange reserves and reduce the current account deficit (CAD). Aggarwal suggested that five years from now, similar concerns could emerge around AI usage because the AI bill for Indian firms could explode.

“That bill could eventually become larger than our import bill for crude oil or gold, and that would be a serious problem. We should be prepared for this paradigm shift in trade,” Aggarwal said. “If we do not build our own models, we will lose the AI race.”

There is also a broader macroeconomic concern. India has 1.4 billion people and is one of the most digitised societies in the world. The amount of data generated in the country is arguably greater than that of any other nation. Yet much of that data does not remain in India because most of the applications Indians use are built by Western companies, which then leverage this data to improve their own models.

advertisement

“Data is our superpower, but ironically, we are not using it for our own advancement. Instead, we are willingly exporting it to enhance the capabilities of Big Tech,” Aggarwal said. “This is a form of modern-day digital extraction, and we need to break this cycle of digital colonisation.”

That is because even today, if an Indian company wants to build a competing application, it often lacks access to the same scale of user data that global technology companies already possess. AI is also not inherently inclusive. While many AI services are free, dependence on them could become permanent as monetisation deepens and paid versions become the norm, widening the gap between who gets to use AI and who doesn’t.

As a result, India may need to move beyond simply imitating Western models and instead create its own innovation cycles. The goal should be to develop indigenous foundation models. And, according to Aggarwal, these cannot merely be small language models; India will also need to build LLMs. There is simply no escaping that reality.

advertisement

Subscribe to India Today Magazine

- Ends