India, Oct. 1 -- As Artificial Intelligence (AI) continues to transform and revolutionise the way we live and work, it's essential to recognise its vast potential and harness its power to drive positive and constructive societal change. It is imperative for India to develop sovereign capabilities in this field for three primary reasons. First, aatmanirbharta (self-reliance) will spur our domestic start-up ecosystem and attract private sector investment. Other AI leaders are already seeing the benefits of this independence. For instance, China recently banned Nvidia chips to boost sovereign chip-making capacity. Second, building AI attuned to the nuances, culture, history, and languages of India will make AI accessible and applicable to every Indian. Third, responsible sovereign capability in this field will become increasingly important for national security. For this, you need advanced models that keep clear records of where their data comes from (data-lineages) to prevent hidden malicious behaviour that could emerge at critical moments. The essential ingredients for building this capability are threefold: Data, computing power, and skilled talent. Today, OpenAI's ChatGPT in India reportedly has more monthly active users than in any other country. OpenAI, which has raised around $40 billion in total funding, is burning $1-2 billion each month to attract users by giving services at zero cost. This predatory pricing is encouraged by US policies, which seek to import data and actively export AI. Apart from user acquisition, companies are using our data to train more superior, closed-source AI models. If India doesn't develop aatmanirbharta in AI, our future may involve having AI services powered by our own data, but owned by others, and then sold back to us. Instead, we should adopt the best features of global models, let them work within India, but set rules that encourage Indian and foreign investment. The Union government, through initiatives like the India AI Mission, has taken bold steps to support Indian start-ups to build sovereign models from scratch. However, unless these models are used on Indian data in the service of every Indian, their true potential will never be realised. India should ensure that sovereign AI reaches the last mile and provides a bulwark against AI use that leads to wealth centralisation. In technology, sometimes those who come second can actually build faster, learn smarter, and skip the pitfalls. A massive opportunity exists for India to leapfrog in AI, by addressing the following issues. First, the government should only deploy sovereign or open-source models that are securely run on sovereign infrastructure for its internal operations. Using closed-source models with non-auditable behaviour poses serious security risks. Additionally, partnerships such as the recent All India Council for Technical Education (AICTE) initiative offering free licences to foreign model providers for students and teachers should be disallowed. Such arrangements result in valuable data leaving India, which can be used to profile Indian citizens and improve foreign models. They also undermine fair competition through predatory pricing. In AI, data is a fundamental input that directly influences the quality and capabilities of the core product. Second, the government should require that all AI models and application programming interfaces (APIs) serving users in India operate on infrastructure hosted entirely within Indian borders. No user data should flow outside India for applications involving large language models (LLMs). This approach will attract significant private investment in computing infrastructure, such as graphic processing units (GPUs) and other accelerated hardware, and improve privacy protections for Indian citizens by preventing data from leaving the country and benefiting foreign AI capabilities at India's expense. Our computing capacity remains vastly behind global leaders. For example, OpenAI and Nvidia will soon be scaling to 10 gigawatts of GPU capacity, whereas India's current capacity is around 30,000 GPUs. Bridging this gap requires massive private sector investment and foreign direct investment (FDI). The government must aggregate demand and provide access to sovereign models as a public good. The same computing infrastructure used for inference can also be harnessed to train new models. This concept aligns with Nvidia CEO Jensen Huang's idea of "intelligence factories" dedicated to AI "manufacturing". The "manufacturing" of AI within India must be promoted with 100% local value addition, while welcoming global capital and intellectual property. Third, the government should work on creating a new type of Digital Public Infrastructure (DPI) that integrates all government services, tourism data, NCERT educational content, emergency services, schemes, railway bookings, the Swayam platform, and similar systems. The government should then encourage developers of sovereign models and applications to create platforms that comply with this infrastructure. These platforms would give all Indian citizens easy access to sovereign AI models, government services, and knowledge bases like NCERT textbooks. This approach would enable any citizen to access a wide range of services through a simple, intuitive platform where they can just ask for what they need. To increase reach, multiple modalities such as voice, mobile apps, and messaging platforms should be used. Fourth, by significantly increasing AI demand in a country of 1.4 billion, both through mandating inference within the country and providing AI services to Indian citizens, India can attract substantial private sector investment in computing infrastructure. This will also encourage FDI and help bring back top global AI talent. Better access to computing resources will reassure talented researchers worldwide that India is a viable place to pursue advanced AI research. India needs a dedicated national programme focused on attracting and retaining top-tier AI experts. This will lead to better methods for curating data and sharing non-personally identifiable information as a public good, as platforms like AI Kosh under the India AI mission are already doing. Ultimately, this will create a virtuous cycle of data, computing power, and talent - driving India's AI capabilities forward and enabling us to leapfrog in the field....