Within two years of its inception, Sarvam AI—founded by Pratyush Kumar and Vivek Raghavan—has grown into more than just a startup. It’s fast emerging as India’s answer to the global AI hysteria. To make generative AI accessible at scale, Sarvam AI is leading one of India’s most ambitious technology missions: developing a sovereign, multilingual, open-weight large language model that can rival global giants. Its core differentiators—openness, affordability, and Indian oversight—set it apart from models such as China’s DeepSeek. “We’re not in a race to copy them,” says Kumar. “There’s little transparency in what they’ve used or achieved. We’ll build something competitive—and make it public.” Kumar is now focused on building a foundational model under the ₹10,000-crore IndiaAI Mission. In May 2025, Sarvam AI launched Sarvam-M, a 24-billion-parameter open-source model, built atop French startup Mistral’s base model and fine-tuned for math, programming, and 10 Indian languages. Though early responses have been measured, Kumar remains undeterred. His message is clear: this is only the beginning. Backed by $41 million from marquee venture capitalists, Sarvam is scaling at breakneck speed. Importantly, while the model may use public datasets, no personal or sensitive information—such as Aadhaar or private data—will be included. “Privacy is non-negotiable,” Kumar insists. If successful, Sarvam could redefine what a truly sovereign AI should look like—for India, and perhaps the world.