To combat bias, the government invested in domestic AI models: MeitY Secy

India Investing in Sovereign AI Models to Reduce Bias and Retain Control: MeitY Secretary S Krishnan

The government is investing in indigenous and sovereign artificial intelligence models to address bias risks and ensure that core AI systems are trained primarily on Indian data and languages, a senior official from the Ministry of Electronics and Information Technology said on Tuesday.

Speaking at a session during the AI Impact Summit, MeitY Secretary S Krishnan stressed that technological sovereignty is a critical requirement for a country of India’s scale and diversity.

“Sovereignty is a very important issue. We need to have AI models where nobody else has the kill switch, and it is always with us,” he said, underlining the need for domestic control over how AI models are built, governed, and deployed.

Krishnan said there was initially a debate over whether India should build its own AI models or rely on open-source models developed abroad, given the higher cost and time required to create indigenous systems. After deliberations, the government chose to proceed with sovereign model development for strategic and social reasons.

The primary motivation, he said, is to reduce bias and improve representational accuracy by training models on Indian datasets across multiple languages and contexts.

“First and foremost is to address the issue of bias. Making sure that we have models which are based primarily on Indian data and data created in a multiplicity of Indian languages — that is very critical,” Krishnan said.

He added that linguistic diversity is another major driver, as users prefer interacting with AI systems in their native languages. This makes domestically built voice-first and voice-to-text AI systems especially important.

India’s flagship sovereign model initiative, BharatGen, was launched in October 2024. The program focuses on building foundational AI capabilities for Indian languages, including Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) models to support multilingual access.

Krishnan clarified that the government’s objective is not necessarily to compete in building the largest general-purpose large language models, but to develop practical AI systems tailored to real-world Indian needs.

“The aim is to create things which can be used by people in their own language and in areas such as science, agriculture, and manufacturing,” he said, adding that sustained public funding is essential to achieve these goals.