India is not looking to slow a frontier AI race in which it is not yet a leading player. Its priority remains building the infrastructure, models and talent needed to strengthen its position in the AI economy. That is why the government has drawn a distinction between calls to slow frontier AI and India’s own focus on applications and models designed for its population.
Why India sees little reason to pause
India’s Ministry of Electronics and Information Technology (MeitY) has said there is no reason for the country to pause AI research because India is not at the frontier end of the technology. The ministry says that much of India’s work is focused on applications and models that can serve domestic requirements, rather than directly competing with the most advanced global models.
The government has not dismissed the safety concerns behind the global debate. MeitY sources said AI-related cybersecurity incidents are already covered by existing reporting obligations, while norms governing the timeframe and content of incident reporting are being tightened. IT Minister Ashwini Vaishnaw has also said the government is working with industry on an appropriate approach to AI regulation and user safety.
A pause aimed at frontier-model development in the US does not necessarily translate into a reason for India to stop building applications, models and infrastructure. For India, slowing development could also mean delaying efforts to build capabilities that are still emerging.
India’s AI race is not only about bigger models
The government’s approach reflects this difference. Under the IndiaAI Mission, 20 indigenous sovereign AI model proposals have been identified for support, comprising 12 large language models and eight small language models. The projects include Sarvam’s 30-billion and 105-billion parameter models, Gnani.ai’s speech-to-speech model, BharatGen’s multilingual foundation models and Avataar AI’s video-generation model. Other projects are focused on healthcare, language technologies and agentic AI.
The mission is also focused on datasets, talent, applications and affordable compute, rather than models alone. By July, the government said it had approved 237 projects, allocated 9.3 million GPU hours and empanelled 15 compute service providers.
That makes India’s AI strategy different from a simple attempt to replicate the US race for the largest model. At Times of India’s AI Quotient Awards 2026, Asia MD Sandip Patel has argued that the more relevant objective for India is to build AI that is affordable, useful, secure and suited to specific requirements, rather than simply pursuing larger models. He has also linked this to control over data and technology infrastructure.
For India, the focus could be on building AI for its own needs, from Indian languages and public services to businesses, healthcare and other key sectors, instead of competing only on model size.
The China question complicates any slowdown
The global debate over slowing AI development also has a geopolitical dimension. The US has opposed calls for a broad halt, arguing that slowing American AI development could weaken its position and give China an advantage. This makes any global slowdown difficult to separate from the wider technology competition between Washington and Beijing.
China, meanwhile, continues to build its AI models, computing infrastructure and domestic ecosystem. For India, this creates a different challenge. While the US and China are already competing at the frontier, India is still building its own compute capacity, indigenous models, talent and applications. A slowdown could therefore have a different meaning for India, which is still trying to close gaps in the AI stack.
India also has to consider its dependence on foreign technology, including advanced models and computing infrastructure. This makes continued investment in domestic capabilities important even as the country works on safety and governance.
For India, the focus could therefore be on continuing AI development while building safeguards alongside it. The objective would not be to match every frontier model, but to strengthen the capabilities India needs to compete, deploy and secure AI as the technology evolves.
The defence layer cannot wait
This is where cybersecurity becomes central to the discussion. Sarvam co-founder Pratyush Kumar has argued that India should accelerate AI development while building its own defensive capabilities. Speaking at the Moneycontrol Startup Conclave, Kumar said increasingly capable AI systems could be used by rogue actors and described AI as a “weapon system”. He called for a domestic defence layer and said cybersecurity should become a priority for governments and companies across the ecosystem.
The argument changes the meaning of acceleration. It is not limited to building more capable models faster. It is also concerned about ensuring that India has the capability to detect, test and defend against increasingly capable AI systems.
Kumar’s comments come as the government develops an AI safety and governance framework alongside the IndiaAI Mission. The framework includes proposed institutional mechanisms such as an AI Safety Institute, while the mission’s Safe and Trusted AI pillar covers areas including risk assessment, privacy-preserving technologies, explainability and bias mitigation.
Whereas, according to Zoho co-founder Sridhar Vembu, India should not wait for a global consensus on slowing AI development. He said a worldwide pause is unlikely and argued that India should keep its strategic options open rather than get locked into the US China technology rivalry. Instead of trying to match the tens of billions of dollars being spent on frontier models, Vembu said India should invest in foundational research and development across sectors, which he argued can build long term technological capabilities without requiring frontier scale spending.
That creates a different way of looking at the global slowdown debate. If more advanced AI systems are likely to create new security risks, waiting for those systems to arrive before developing the defensive layer could leave India dependent on capabilities developed elsewhere.
A window to build, not to catch up
India’s AI development is not simply a choice between moving faster or slowing down. The country is still building the infrastructure needed to participate in the AI economy. Under the IndiaAI Mission, the government is expanding shared compute capacity and supporting indigenous models and applications.
Indian companies are also looking beyond applications. Sarvam has said it plans to develop a trillion parameter model, showing that some Indian developers are also looking to compete at the foundation model level.
But continuing AI development does not mean replicating Silicon Valley’s race to build ever larger models. India’s more immediate challenge is to strengthen the wider ecosystem around AI, including compute, talent, applications and security.
Dependence on foreign AI providers adds another concern. When some AI entities restricted access to some of its models in India, Activate founder Aakrit Vaish told TechCrunch that the episode could push Indian startups towards open source models and reduce reliance on a small number of foreign providers.
For India, a period of slower frontier development could therefore be used to address gaps at home, from expanding compute and developing models for local requirements to strengthening the security systems needed to protect AI deployments.
