Krutrim, the AI startup founded by Ola’s Bhavish Aggarwal, has launched what it calls India’s first ‘sovereign’ large language model built for enterprise workloads. The model, Krutrim Enterprise 2.0, supports 22 Indian languages and has been fine-tuned for compliance-heavy sectors like banking, insurance, and government services.

Unlike OpenAI’s GPT or Google’s Gemini, Krutrim’s model is trained entirely on Indian data and hosted on Indian cloud infrastructure. The company claims this is critical for enterprises dealing with sensitive citizen data that cannot leave the country under India’s evolving data protection framework.

The model is available via API and through a managed enterprise platform that includes guardrails, audit logging, and role-based access control. Early customers include two public sector banks, a state government, and three insurance companies.

“Enterprise AI in India cannot depend on American cloud providers for everything,” said the company’s CTO during the launch event. “Data sovereignty isn’t just a buzzword — it’s a regulatory requirement that’s coming fast.”

Krutrim raised $50 million in its seed round in early 2024, making it one of India’s fastest startups to reach unicorn status. The company is currently in talks for a Series A round at a reported $3 billion valuation.