SEBI's Coming AI Rules: Kill Switches, Human Oversight and No Dilution of Accountability

SEBI chairman Tuhin Kanta Pandey said on Saturday, 10 October, that the regulator will shortly issue guidelines on the responsible use of AI and machine learning in India's capital markets. He was speaking at the second Capital Market Confluence 2026, organised by the Bombay Stock Exchange Brokers' Forum.

What the framework will contain

The framework will be tiered, Pandey said, with clear accountability, data controls, kill switches and human oversight built in. It will also follow the International Organization of Securities Commissions' supervisory toolkit for AI, so that governance can keep pace as the technology changes.

"Technology cannot dilute accountability," he said. Regulated entities stay responsible for AI outcomes, data quality, model governance and cyber resilience, and human oversight has to remain clear even as AI changes how markets operate and how they are supervised.

The rest of the agenda

AI was one item on a longer list.

SEBI expects to issue guidelines on the settlement price of derivatives on expiry days in about a week, after reviewing comments on a consultation paper that followed its review of the Closing Auction Session. It is targeting the end of November 2026 to bring tender offers, buybacks and offers for sale into the interoperability framework, and it has received more than 400 comments on simpler digital onboarding for Persons Resident Outside India.

For scale: India's market capitalisation is around Rs 472 lakh crore, with more than 15 crore unique investors and mutual fund assets of about Rs 87 lakh crore.

Why this matters for brokers, AMCs and fintechs

For any broker, AMC or market intermediary running algorithmic or ML-driven systems, a kill switch and a named human owner come down to being able to show who owns each model, what data it was trained on, and how fast it can be stopped. A tiered design suggests the bar will rise with risk. A customer-service chatbot and an order-routing model are unlikely to be held to the same standard.

Firms that already keep a model inventory and a clear approval trail will find this easier than those that have let AI tools spread team by team. The guidelines aren't out yet. Getting the inventory in order before they land will cost less than doing it after.

Source: The Tribune (ANI) — SEBI to issue AI, machine learning guidelines shortly