

CHENNAI: India should pursue an artificial intelligence strategy centred on smaller, specialised and resource-efficient systems rather than compete in the race to build increasingly large and expensive models, researchers at the Indian Institute of Technology Madras (IIT-Madras) have proposed in a new white paper.
Titled ‘A Frugal AI Framework for India’, the paper advocates ‘Responsible Frugal AI’, an approach that seeks to maximise public value while minimising computational requirements, energy consumption and financial costs without compromising reliability, robustness or accountability.
Released by IIT-Madras Director V Kamakoti at the Industry Conclave on AI Governance 2026, the paper presents an alternative to replicating the capital-intensive AI development models pursued by countries such as the US and China or relying on foreign technologies.
“AI does not have to know everything to be useful. It has to know the right things, deeply, for the task at hand,” Kamakoti said, emphasising the need for affordable, effective and sustainable solutions.
The paper was authored by researchers from the institute’s Centre for Responsible AI (CeRAI), Wadhwani School of Data Science and AI (WSAI), and Department of Engineering Design.
CeRAI and WSAI head Balaraman Ravindran said India could chart its own course by leveraging its strengths in engineering, innovation and digital public infrastructure.
CeRAI Research Director Gokul S Krishnan said institutions should first establish whether AI was necessary for a particular task and then identify the smallest model capable of delivering the desired outcome. Simpler alternatives, including rules-based systems, retrieval-based approaches and conventional machine learning, should be considered wherever adequate, he said.
The framework calls for shared AI infrastructure, modular and interoperable systems, and multilingual applications designed for India’s diverse population. It also recommends selective deployment of generative AI and evaluating systems on practical utility, resource consumption, accountability and social value rather than benchmark performance.
A seven-point checklist urges institutions to justify AI adoption, define operational constraints, select appropriate models, conduct real-world testing, monitor lifecycle costs and resource consumption, and ensure documentation, independent audits, human oversight and manual alternatives. It also recommends publishing deployment failures to improve accountability and future implementation.