Inside Sebi's Growing Use of AI to Catch Market Manipulation and Fraud

Inside the Securities and Exchange Board of India's complex in Mumbai's financial district, analysts are increasingly relying on artificial intelligence to keep pace with some of the world's most sophisticated traders, according to people familiar with the regulator's operations.

How the system works

Sebi's surveillance teams use AI tools to sift through streams of trading data — stock volumes, derivative positions and algorithmic identifiers — looking for anomalies that might signal manipulation. The system also tracks suspicious stock tips and promotional content across social media and online forums, cross-referencing them against actual trading activity to spot potential pump-and-dump schemes.

One recent case illustrates the speed AI has brought to enforcement. Last month, Sebi took enforcement action against a JPMorgan Chase unit just six days after its surveillance systems flagged suspicious activity in a newly launched closing auction mechanism. An alert generated by the regulator's machine-learning tools was among the factors that underpinned the case. The regulator has since lifted the trading ban after the unit returned the alleged illegal gains, though the broader investigation continues.

Built over several years

Sebi's AI push traces back to 2019, when it began investing in a local data centre at its Bandra Kurla Complex headquarters, which went live in 2021 — the same year the regulator started deploying its in-house AI tool, named Sudarshan. Early versions reportedly struggled with hallucination-prone large language models, but the system has since matured considerably. Sebi's technology team has grown to more than 200 engineers and analysts, up from roughly 20 just five years ago.

The tools have also sped up routine regulatory work: machine learning has reportedly cut review times for initial public offering filings by as much as 70%, in addition to generating more precisely targeted alerts for unusual trading patterns.

Tackling social media manipulation

A persistent challenge for Sebi is misinformation aimed at India's roughly 140 million retail investors. The regulator now sends as many as 7,000 monthly requests to platforms including X, Instagram and Telegram to take down misleading content — a roughly 40% increase from pandemic-era levels — and says its systems have helped remove more than 100,000 videos across these networks.

The urgency behind this build-out was sharpened by the Jane Street episode last year, when Sebi accused the quantitative trading firm of manipulating the Nifty Bank index, temporarily barring it from the market and ordering the return of alleged illegal gains. Jane Street has denied wrongdoing, and its trading ban was lifted after it deposited the disputed funds, but the case exposed how complex modern trading strategies have become for regulators to police manually.

Not without limits

Industry voices caution that AI cannot fully replace human judgement in enforcement. "AI needs to be seen as an assistant, not as an autonomous regulator," said one AI-focused industry expert, noting that Sebi officials still review AI-generated findings before acting to ensure decisions hold up to legal scrutiny. Challenges remain too, including the risk of "data poisoning," where manipulated inputs — including bot-generated social media content — could distort a model's outputs, and constraints on computing power amid a global shortage of graphics processing units.

Sebi did not respond to requests for comment on its AI programme, but the direction of travel is clear: as India's derivatives market has grown to nearly $4 trillion in notional value, technology has become as central to market oversight as it is to the trading strategies it is meant to police.

This is an original summary based on public reporting. See our editorial policy for how we source, write, and correct our stories.