‘AI is a weapon system now’: Sarvam’s Pratyush Kumar says India needs its own defence layer

New business data points to the fact that India needs to build its own defensive capabilities around artificial intelligence as the technology becomes more powerful and potentially available to rogue actors, Sarvam co-founder Pratyush Kumar stated in conversation with Nalin Mehta, Managing Editor of Moneycontrol, at the Moneycontrol Startup Conclave 2026 on Saturday.

“The immediate consequence of AI is usage of it by rogue players. This is a weapon system now,” Kumar stated during a session titled India’s AI Unicorns: Models, Moats & the Race to Scale.

“We need a defence layer. We absolutely need to be thinking around our own capabilities,” he stated.

Kumar, who was speaking alongside Emergent founder Mukund Jha, stated the risks surrounding AI should not translate into India slowing its development of the technology.

“From the India vantage point, we should accelerate the pace of AI,” he stated.

By The Way | Not just Sarvam models: Indian AI startup can now host American models too

At the same time, Kumar stated governments and firms would have to consider the vulnerabilities created by increasingly capable systems.

“From a national interests perspective, we need to think of vulnerability. It can go the other way if not regulated,” he stated.

Cybersecurity, he further noted, would have to become a priority across the ecosystem.

“Whether it is government, startups, everybody needs to take cybersecurity seriously,” Kumar stated.

He additionally pointed to the changing capabilities of AI systems while discussing the difficulty of managing the pace at which the technology is developing.

“Pacing the AI has multiple layers,” Kumar stated. “Technologically, it is crazy with models deciding amongst themselves.”

Do Not Miss | From AI models to agents: Sarvam’s Kumar, Emergent’s Jha map India’s next five years

India builds AI safety framework

Kumar’s comments come as the Centre builds out a safety and governance framework alongside the IndiaAI Mission.

The government’s AI Governance Guidelines, released in 2026, proposed institutional mechanisms including an AI Governance Group, a Technology and Policy Expert Committee and an AI Safety Institute. The framework is intended to address emerging risks while supporting the development and adoption of AI.

The Safe and Trusted AI pillar of the IndiaAI Mission includes work on machine unlearning, bias mitigation, privacy-preserving technologies, explainability and risk assessment. Thirteen responsible AI projects had been selected under the programme by July 2026, according to the government.

Sarvam itself has additionally identified cybersecurity as one of the areas for further model development. The firm stated in June that funding from its Series B round would be used, among other areas, for research on its next frontier model for agentic, coding and cybersecurity use cases.

Sarvam was selected under the IndiaAI Mission in April 2025 to build a sovereign large language model ecosystem. The government stated the project would involve an open-source model trained for Indian use cases, including governance and public-service applications.

Kumar’s remarks at the conclave placed cybersecurity and domestic capability alongside that effort to expand India’s AI capacity.