Tuesday, November 17, 2026 | 07:17 PM ISTहिंदी में पढें
Business Standard
Notification Icon
userprofile IconSearch

Foreign AI models can turn India into a digital colony: Sarvam co-founder

Sarvam AI co-founder Vivek Raghavan warns against dependence on foreign AI models, calls for sovereign technology and stresses the need for Indic language and voice-first AI

Vivek Raghavan, Co-Founder of Sarvam
premium

Vivek Raghavan, co-founder of Sarvam AI

Avik Das Bengaluru

Listen to This Article

India risks becoming a digital colony if it continues to rely on foreign artificial intelligence (AI) models by renting the stack rather than building sovereign models of its own, Vivek Raghavan, co-founder of Sarvam AI, said.
 
He said that when a user submits a query to an AI system, the data leaves the country and is used to train models abroad, making them more nuanced, responsive and better, leading to greater usage.
 
“A lot of people are proud to say that India is largest user of ChatGPT. But that is not a good thing. We are exporting data and importing intelligence. A 100 years ago, we were exporting cotton and importing cloth and that was a colony and there is a danger we may become a digital colony,” he said during a keynote presentation at the Bengaluru Tech Summit.
 
The topic of sovereign AI has assumed great importance in the last two years as the US and China build overwhelming supremacy in the field with incredibly powerful models, while other countries play a catch-up game and become completely dependent on them. That has raised concerns that the technology, expected to deeply impact lives, will be used by the two superpowers as a bargaining chip that could throttle development elsewhere.
 
Sarvam, the Indian frontier model company, launched a 105-billion-parameter model earlier this year and plans to launch a trillion-parameter model soon, which will be used for coding, cybersecurity, biomolecules and additive manufacturing, which cannot be done with small models.
 
On the adoption of AI in India, Raghavan said four factors determine the scale and pace of adoption. They are Indic languages, voice-first interfaces, low data costs and design for scale.
 
“Most AI is English first and models have become better in Indian languages. But if you want to build AI for India, you have make it native to Indian languages. Almost 57 per cent people prefer to access AI in their own language,” he added.
 
Voice AI has to be the main interface, as espoused by other experts, too. With 22 scheduled languages and 6,000 dialects, it is believed that whoever cracks this model will be able to deliver AI in India at scale.
 
India is expected to be the biggest test bed and opportunity in the sector, with 1.4 billion people and hundreds of millions still using feature phones. Nandan Nilekani, non-executive chairman of Infosys and Aadhaar architect, believes voice AI is the final frontier and the only practical interface to bring about true digital equality in India.
 
Meanwhile, the company, in a blog post on Tuesday, said that it had opened Token Factory to developers and enterprises. This will give them access to Sarvam’s and global AI models through a single application programming interface.
 
Sarvam is positioning Token Factory as an Indian alternative for computing infrastructure to run models on GPU clusters located in India.