New Delhi: In a strategic move to fortify national security and promote technological self-reliance, the Indian government is reportedly planning to mandate the use of domestically developed Artificial Intelligence (AI) models across critical sectors. Sources suggest that key government departments, particularly those operating in defence and aerospace, may soon be directed to completely phase out the usage of advanced AI systems and Large Language Models (LLMs) developed by foreign corporations.
This ambitious transition is being driven by the government’s flagship IndiaAI Mission, which was approved in March 2024 with a substantial budget allocation of Rs 10,372 crore. The primary objective is to minimise reliance on global tech giants and ensure that sensitive governmental operations are powered by secure, homegrown technology. According to senior officials, certain defence units are already constructing highly secure, “air-gapped” AI systems that operate on the edge, entirely disconnected from external networks to prevent data breaches and cyber threats. These departments are now being encouraged to integrate the indigenous AI models being cultivated under the IndiaAI initiative.
Under the massive funding umbrella of the IndiaAI Mission, the government has sanctioned the development of 20 distinct AI models. This comprehensive lineup includes both Large Language Models (LLMs) designed for general-purpose applications and tailored Small Language Models (SLMs) catering to sector-specific requirements. The responsibility of building these foundational models has been entrusted to a curated group of 12 Indian startups.
Prominent names like Sarvam and BharatGen have already made significant strides by unveiling several AI models. Joining them are other homegrown innovators such as Soket, which recently launched its video-generation AI model ‘Varya’ and the foundational model ‘EKA-1’. Additionally, ZenteiQ has introduced its ‘Dhir 2B’ and ‘Dhir 5B’ models. Several other tech firms, including Tech Mahindra, Fractal, Gan AI, Gnani, and Shodh AI, are also rigorously working under this mission and are expected to roll out their respective LLMs and SLMs within the next 12 to 15 months.
Government officials assert that the current crop of domestically produced AI models has already achieved approximately 80 per cent of the capabilities demonstrated by the world’s most advanced AI systems. Through continuous dialogues and partnerships with these selected startups, the government aims to further refine these models, ensuring they are robust enough to handle the complex and highly sensitive demands of India’s critical infrastructure.