SEBI Prepares AI Controls for Indian Markets

Pandey identified opacity, bias, cybersecurity, data protection, and accountability as risks associated with AI deployment in markets, ANI reported. He said AI could strengthen surveillance, risk assessment, fraud detection, and investor services, while arguing that its use needs to preserve market trust.

The proposed requirements would establish an explicit operational-control expectation for regulated entities deploying AI or ML. A kill switch generally provides a means to halt an automated system when its behavior, outputs, or surrounding conditions create an unacceptable risk. Human-in-the-loop controls, meanwhile, reserve specified review or intervention responsibilities for people rather than leaving decisions entirely to automated workflows.

The framework has not yet been published, so technical requirements, covered entities, implementation dates, audit criteria, and enforcement mechanisms remain undisclosed in the reporting. In regulated financial settings, comparable governance regimes commonly turn high-level controls into requirements for model inventories, access management, logging, testing, escalation paths, and documented human authority to override automated outcomes. For ML teams, a kill-switch requirement can therefore extend beyond a simple application toggle: effective implementation typically depends on dependable rollback procedures, observability, ownership, and tested incident-response processes.

ANI reported that Pandey made the announcement while describing rapid growth in India's markets, including equity issuance of more than Rs 4.5 lakh crore in FY25-26 and about 14.9 crore unique investors. The forthcoming SEBI document will be needed to assess the practical compliance burden on regulated entities.