The Securities and Exchange Board of India is building an AI-assisted surveillance system to detect suspicious trading, trace market manipulation across social media and accelerate regulatory investigations. The shift reflects a basic institutional problem: India’s markets are becoming faster, more complex and more heavily influenced by digital information than a regulator relying only on conventional human review can easily monitor.
According to a Bloomberg News report carried by NDTV Business, SEBI analysts in Mumbai use artificial intelligence tools to examine stock volumes, derivatives positions and algorithmic identifiers for unusual patterns. The system also tracks stock tips and promotional posts on social media and online forums, matching them with trading activity to identify possible pump-and-dump schemes.
The reported system is not described as an autonomous enforcement mechanism. Officials continue to review AI-generated alerts and adjust the outputs so that any regulatory action complies with existing rules and can withstand legal scrutiny. That distinction is central to understanding SEBI’s technology push. The regulator is not handing enforcement to a machine; it is using machine-assisted detection to decide where scarce investigative attention should be directed.
The immediate pressure comes from the scale and speed of India’s securities markets. The report describes nearly $4 trillion in notional derivatives trades, sophisticated high-speed trading strategies and a large volume of social-media commentary and misinformation. These conditions create a surveillance challenge that is not simply about processing more data. It is about linking different kinds of data quickly enough to identify relationships that may be invisible when each stream is examined separately.
A recent enforcement action involving a JPMorgan Chase & Co. unit illustrates how the system may operate in practice. According to a person familiar with the matter, SEBI initiated action six days after surveillance systems flagged suspicious activity connected with a newly launched closing auction mechanism. An AI and machine-learning alert was among the factors supporting the regulator’s case. SEBI later lifted a trading ban after impounding alleged illegal gains, while the investigation continued.
The account is significant not because it proves that AI can independently identify wrongdoing, but because it shows how automated alerts are becoming part of the evidence-gathering chain. The technology can narrow the time between an unusual transaction and an initial regulatory response. That matters in markets where trading strategies can change rapidly and where evidence may become harder to reconstruct after the event.
The Jane Street episode added urgency to this effort. In July 2025, SEBI accused the quantitative trading firm of manipulating the Nifty Bank Index, temporarily barred it from the market and ordered it to return billions of rupees in allegedly illicit gains. Jane Street denied the allegations, and the ban was later lifted after the firm deposited the alleged illegal gain. The episode underscored the difficulty of assessing complex trading strategies and the consequences of regulatory intervention in markets that attract sophisticated participants from around the world.
SEBI’s response has involved more than purchasing software. The regulator has expanded its team of engineers and analysts to more than 200 employees from about 20 five years ago, according to people familiar with the matter. Machine learning has reportedly helped generate more targeted alerts for unusual activity and speed up parts of the review process for initial public offerings by as much as 70 per cent.
That staffing growth is important because effective financial surveillance depends on institutional capacity as much as computing power. A model can identify an anomaly, but officials must determine whether it reflects manipulation, a legitimate trading strategy, a data error or activity that requires further evidence. The human team provides the legal, market and procedural interpretation that turns a technical signal into an administratively usable case.
SEBI’s technology programme also extends beyond trading screens. The regulator is reported to make as many as 7,000 requests a month to platforms including X, Instagram and Telegram to remove misinformation, around 40 per cent higher than pandemic-era levels. More than 100,000 videos have reportedly been taken down across networks.
This connects market surveillance with a broader urban and institutional reality: financial regulation now operates inside the same digital environment as retail investors, influencers, messaging groups and automated accounts. The regulator’s challenge is not limited to exchanges and brokerages. It also includes the information networks through which ordinary investors receive tips, warnings and promotional claims.
The scale of India’s retail investor base makes this particularly consequential. The report puts the number of retail investors exposed to such information at 140 million. A misleading post can therefore become more than an isolated piece of bad advice. When combined with coordinated trading activity, it can become part of a market event affecting large numbers of households.
SEBI’s AI push has been developing for several years. In 2019, it began investing in a local data centre at its headquarters in Mumbai’s Bandra Kurla Complex. The facility became operational in 2021, the same year the regulator began deploying its in-house AI tool, Sudarshan. Early versions reportedly struggled with large language models that produced hallucinations, or plausible but incorrect outputs.
That history shows why regulatory AI cannot be judged only by its speed. Financial enforcement requires traceable evidence, consistent procedures and decisions that can be explained to affected parties. A model that generates a useful lead but cannot show how it reached that result may assist an investigation while still being unsuitable as the sole basis for punitive action.
The risks identified in the report are therefore institutional, not merely technical. Data poisoning could allow manipulated training data to alter a model’s behaviour. Social-media material may include bots or coordinated false content, creating the possibility that the system treats malicious inputs as evidence. SEBI’s in-house systems also face scaling constraints because access to graphics-processing units remains limited by a global supply shortage.
These problems are amplified when regulators use external data sources. A social-media post may be relevant, but its origin, authenticity and relationship to a trade must be established separately. Automated systems can help connect the post to market activity, but they cannot by themselves establish intent or legal liability. That requires a process in which data provenance, model performance and human decisions remain reviewable.
The governance model described by experts in the report is consequently one of assisted regulation. Pankit Desai of Sequretek said AI could allow a genuine case to be built within hours or days rather than years. JP Mishra of Deep Algorithms Solutions, however, said AI should be treated as an assistant and not an autonomous regulator. Rishi Kapoor of Asifma similarly argued that regulators must apply strong data confidentiality, model governance and risk-management standards to their own systems.
This creates a parallel with the obligations SEBI imposes on market participants. Firms are expected to manage operational risk, preserve records and explain their conduct to the regulator. As SEBI expands its own use of automated systems, it faces a comparable need to document how models are trained, how alerts are prioritised and how human officials validate the results.
The policy landscape is still developing across Asia, where financial regulators are at different stages of adopting AI internally. The available account does not establish a common regional framework or specify new rules that SEBI has formally adopted. It does show, however, that the regulator’s technology programme is moving from experimentation towards operational use in surveillance, investigations, IPO review and online misinformation response.
The larger urban question is one of institutional infrastructure. SEBI’s data centre, engineering workforce, computing capacity and relationships with digital platforms form a regulatory system embedded in Mumbai’s financial and technological ecosystem. This infrastructure is less visible than a metro line or a public building, but it increasingly shapes how the city’s financial economy is governed.
The evidence confirms that SEBI is using AI to improve the speed and targeting of market surveillance, while retaining officials in the decision-making loop. It also confirms that the system faces risks involving unreliable data, model errors, limited computing capacity and legal accountability. The developments that require monitoring are therefore not only the next enforcement action, but also how SEBI documents model governance, protects confidential data and demonstrates that automated alerts can support fair and defensible regulation.
