The Indian Institute of Technology Madras (IIT Madras) will host an industry conclave on AI governance 2026 on October 22, bringing together researchers, industry representatives and policy experts to examine how artificial intelligence (AI) systems can be evaluated and governed in real-world settings.
Organised by the Centre for Responsible AI (CeRAI) at the Wadhwani School of Data Science and AI (WSAI), IIT Madras, the conclave will focus on the theme ‘AI Measurement and Governance’, with particular attention to India and the Global South.
The event comes amid the rapid adoption of AI across industries, even as policies and regulatory frameworks governing its deployment continue to evolve. A key focus will be on developing reliable measurement and evaluation methods to ensure that AI systems are fair, safe, robust and accountable in the contexts in which they are used.
IIT Madras Head, WSAI and CeRAI, Prof. Balaraman Ravindran, said governance must be supported by measurable evidence of how AI systems perform.
“Governance that isn’t grounded in measurement remains a statement of intent, not a practice. If we want AI systems to be fair, safe, and robust, we need to be able to demonstrate its evaluation to be able to govern. That is the gap this conclave is designed to address,” he said.
Global experts to deliver keynote addresses
The conclave will feature keynote addresses by Prof. Gopal Ramchurn, University of Southampton and CEO of Responsible AI UK (RAi UK), and Prof. Mohan Kankanahalli, Provost’s Chair Professor of Computer Science and Director of the NUS AI Institute at the National University of Singapore.
Prof. Ramchurn said the growing use of AI, particularly agentic AI systems capable of operating with limited human supervision, raised important questions about accountability and risk management.
He said the conclave would examine who is responsible for the actions of deployed AI systems, how that responsibility can be traced and how risks arising from unintended or rogue behaviour can be mitigated.
Prof. Kankanahalli highlighted the challenge of evaluating increasingly capable AI models, noting that benchmarks can quickly become obsolete as technology advances. Reliable measurement, he said, remains essential for effective governance.
Focus on consumer rights and AI accountability
A fireside chat on AI measurement and governance will feature Rahul Matthan, Partner, Trilegal, and Prof. Balaraman Ravindran.
The programme will also include a panel discussion on AI and Consumer Rights, bringing together S. Saroja, Executive Director, Citizen Consumer and Civic Action Group; Saranya Gopinath, Head of Policy, Razorpay; Prof. Shouvik Guha, Faculty, National University of Juridical Sciences; and Saksham Malik, Government Affairs and Public Policy Manager, Google India.
The discussion will be moderated by Vibhav Mittal, Partner, Anand and Anand, and Affiliate Research Fellow at CeRAI.
The panel is expected to examine consumer protection concerns arising from AI deployment and the role of governance frameworks in addressing them.
Startups, research tools and workshops
A startup showcase will feature companies working on AI measurement, evaluation, observability, compliance and governance, providing opportunities to engage with researchers and policy experts.
CeRAI researchers will present the centre’s research highlights, evaluation methods and open-source tools for responsible AI development and deployment. Three parallel hands-on workshops will focus on responsible AI tools and frameworks.
The conclave will also mark the launch of a new research initiative and the announcement of a paper, with further details to be disclosed at the event.
As organisations move from AI principles to implementation, the discussions will focus on building practical measurement systems and standards that can support responsible AI adoption across sectors, particularly in settings where social, economic and regulatory conditions differ.
The emphasis on India and the Global South is intended to bring greater attention to the need for AI evaluation methods that reflect local contexts rather than relying exclusively on benchmarks developed elsewhere.