AI in action · Regulation
SEBI Weighs an India-Built AI Model to Supervise Markets
India's securities regulator is exploring an AI model built and governed in India for its market oversight, chairman Tuhin Kanta Pandey said on October 28, citing concerns about access to frontier models. SEBI is also adopting IOSCO's AI toolkit.
The Securities and Exchange Board of India is considering building its own artificial intelligence model for the securities market, chairman Tuhin Kanta Pandey said at the Business Standard BFSI Insight Summit in Mumbai on October 28. The model would be developed and governed in India, and Pandey suggested it could serve as digital public infrastructure for the market, available to the institutions SEBI supervises, Business Standard reported.
His reason was dependence. Pandey cited concerns over the confidentiality and availability of advanced AI models, most of which are developed and hosted abroad. A regulator that relies on a foreign provider's model to analyse trading data has to send that data outside its control and accept that access can be restricted, a concern that has grown since some frontier models were offered only to selected institutions this year. No timeline, budget or technical design was given, and the idea is at an exploratory stage.
The more concrete commitment was to an international framework. SEBI is working to implement the supervisory toolkit for AI published by the International Organization of Securities Commissions in May, which helps regulators assess machine learning, generative AI and agentic systems used by market participants, with attention to governance, transparency, third-party dependencies and black-box models. Pandey said AI tools could flag patterns that need further examination, but that their use depends on reliable data, model oversight and clear accountability, and that faster, automated processes make institutional judgement more important, not less.
The speech also covered the plumbing. SEBI is preparing a technology roadmap for market infrastructure institutions such as exchanges and depositories, has proposed extending its IT and cybersecurity framework to their subsidiaries, and plans a training programme with Rashtriya Raksha University and the National Institute of Securities Markets for security operations engineers on using AI to analyse alerts and automate responses.
India's approach mirrors a debate elsewhere: regulators want the benefits of AI supervision without becoming dependent on a handful of foreign providers for a core public function. For market participants, the IOSCO toolkit is the part that matters sooner, because it sets out what supervisors will ask about their own AI systems, from model inventories to who is accountable when an automated decision goes wrong.
Sources: Business Standard — BFSI Insight Summit: Sebi explores domestic AI model for securities market (Oct 28, 2026); Angel One — SEBI considers developing India-specific AI model for securities market regulation; IndiaIPO — SEBI to implement global supervisory toolkit for AI: Pandey