[Courting Controversy] AI Roulette: Can SEBI keep the chamber empty?

“It is madness to risk what you have and need for what you don't need.”

Warren Buffett, the Oracle of Omaha, has long cautioned against investment vehicles that offer a usually win, occasionally die payoff, likening excessive risk-taking to a game of Russian Roulette: the odds may favour you most of the time, but any one bad outcome can be catastrophic. Buffett’s warning was meant for human decision-makers, who could at least be reasoned with, restrained, or held accountable for their choices. There is, however, opacity to the risk when Artificial Intelligence (“AI”) and Machine Learning (“ML”) enter the picture. 

The use of AI in the stock market was recently highlighted at the 23rd FICCI Capital Markets Conference 2026, when the SEBI Chairman Mr. Tuhin Kanta Pandey, declared that SEBI was preparing guidelines for the responsible use of AI and ML in the securities market. The proposed framework is expected to follow a tiered approach, calibrating requirements to the risk posed by a given use case, and will include human oversight requirements, stronger data controls, and kill-switch mechanisms that allow AI systems to be halted quickly if they begin to behave abnormally.

However, this is not a cold start. SEBI has been alive to the use of AI by market players and has been working towards designing a comprehensive framework which governs different aspects of such use. In May last year, SEBI came out with a circular for reporting AI and ML applications offered and used by Mutual Funds (“Mutual Funds Circular”). In order to gain an in-depth understanding of the AI and ML technologies, the Mutual Funds Circular set in place a reporting mechanism for mutual funds that were using AI or ML applications as offerings to investors or internally in their operations or to disseminate investment strategies or advice or to carry out compliance. The reporting mechanism involved detailing how an AI application was implemented and what safeguard mechanisms were in place to prevent abnormal behaviour by such applications.

This graded response from a narrow reporting mandate for mutual funds to a possible market-wide framework covering oversight, data controls, and kill-switches is only possible because of the kind of powers SEBI possesses. SEBI’s broad normative function allows it to formulate rules and regulations and issue circulars or advisories on matters affecting the securities market, with an objective of maintaining order in the securities market and protecting investor wealth. In fact, the Supreme Court has categorically held that SEBI’s wide powers, coupled with its expertise and robust information-gathering mechanism, lend a high level of credibility to its decisions as a regulatory, adjudicatory and prosecuting agency.

The space to watch is going to be whether SEBI’s regulations would be able to keep pace with the advancement of AI. For one, on what basis can AI be held to be accountable for advice that may be provided or investments it may encourage. For example, AI and ML add a layer of opacity to identifying risk in decision making. It is this gap between AI’s growing role in investment decisions and the market’s ability to understand, monitor, and, if needed, stop that decision-making, that the Indian securities regulation needs to confront.

In fact, the mechanism set out in the Mutual Fund Circular are a step in covering such opacity by requiring disclosures regarding explainability of the AI models in use. However, it is unclear that apart from it being an information gathering exercise, how SEBI intends to use its powers if it is dissatisfied with that disclosure. There is also no threshold set out in place on the adequacy of the logic and information used by the AI software in giving advice.

Fortunately, SEBI’s wide powers which are normative, executive, and adjudicatory in nature, equip it to do so. No one is likely to argue that the answer to the problem is a complete prohibition on the use of AI.  The fact remains that AI offers real gains in efficiency and fraud detection; some of which are being harnessed by SEBI itself through its initiatives such as Project SUDARSAN, an AI-based surveillance tool that scans social media to flag fraudulent investment content and impersonators posing as registered advisors, and R(AI)DAR, which reviews mutual fund advertisements and investor-education material for undisclosed or misleading claims.

The proposed guidelines would be an opportunity for SEBI to include clearly defined human oversight requirements for consequential decisions, explainability standards to avoid evasion of scrutiny, circuit-breaker protocols that are tested periodically and a liability framework that assigns responsibility regardless of whether the technology was built in-house or procured externally.

As AI becomes further embedded in investment advisory and compliance, a proactive approach will allow India’s markets to absorb AI’s benefits while reducing its risks. If AI keeps the game of Russian Roulette in play, SEBI is what ensures the gun stays unloaded. It remains the institution best placed to make that call.

Nakul Dewan is a Senior Advocate and King’s Counsel.