India's AI strategy: Balancing innovation, risks, and proper governance

Prime Minister Narendra Modi’s remarks at the global Artificial Intelligence (AI) Action Summit in Paris indicate that India’s policymakers are looking in the right direction when it comes to AI. He identified the threat of job losses as among the biggest concerns, along with concerns about cyber security, disinformation, and deep fakes. As he said, there is a need to reskill populations to align them with the prospects of an AI-driven future. Mr Modi also outlined the need to develop new green-energy sources because the new technologies will increase demands for clean, stable power. He indicated that India’s AI mission would look to build public-private partnerships around key AI resources with the government investing in hardware compute resources and in primary research and development (R&D), and on top of this the private sector may build commercial applications. 

India has one of the world’s largest AI talent pools due to its technically proficient personnel. Its plethora of sophisticated local languages spoken by large populations could also make it a rich ecosystem for research. The government is investing in building large language models, which are likely to be open-source with a low-cost application programming interface, possibly on the lines of DeepSeek. India has also been flagged as one of the largest markets for AI. Sam Altman of OpenAI, who was in India recently, said the country was the second-largest market for ChatGPT, and that’s no surprise, given the large pool of mobile broadband users, rapid growth in startups, and a digital payments system that generates vast data. Translating AI ambition into reality will, however, require a fair amount of agility. The contours of AI are changing rapidly as open source takes hold and new applications are generated. ChatGPT, for instance, launched on the assumption that massive investment would be required to create big proprietary models and that was also the case with Claude, Gemini, Llama, and Grok. But the advent of DeepSeek has turned those assumptions upside down. 

The government’s intentions in terms of developing compute capacity and its focus on encouraging the proliferation of data centres are in the right direction. But it will need to be flexible and highly responsive to changes. India must not get locked into technology or hardware that is left behind by new developments, given the rapid pace of change. Policymakers must also watch out for models that develop a bias as Mr Modi pointed out. AI builds models on available data and it has a tendency to incorporate existing biases. Face-recognition programmes trained on Caucasian images don’t work well with Asian faces, for instance, and any gender, race, or caste biases in the data will tend to be incorporated in the model. While this is a global problem, it could be accentuated in India with its massive disparities. Correcting for bias along with ensuring technology-agnostic development will be critical if AI is to serve Indians well. 

Policymakers also need to be proactive about amending relevant legislation as AI develops new capabilities. New unforeseen issues could develop and require rapid rewriting of laws. One area of concern is the induction of AI into law enforcement and defence. AI can be used as an instrument of repression since it greatly expands the scope of surveillance. It can be a transformational tool in defence applications where it can be a force multiplier in many ways but again, it can cause untold destruction if deployed indiscriminately. Good governance in the age of AI would include ensuring that it is not turned into a tool of repression. Policy at both global and local levels must create conditions for AI to be used in enhancing efficiency and development outcomes.