SEBI Needs AI-Powered Surveillance Upgrade 

With an exponential growth in investor participation and rising trading activity across segments, the market regulator, Securities and Exchange Board of India, faces a challenging task in ensuring orderly functioning of the securities markets with integrity. The highly digitised order flows and complex modus operandi adopted in market abuse, manipulation, price rigging, front-running, and fraudulent trading by a few motivated participant groups make it difficult for the regulator to detect such violations at early stages.

The market surveillance function in the Indian markets runs at two levels – i.e., the individual exchange and the regulator. Apart from low integration of their systems, the inadequacy of organisational capabilities, resources, procedures, and tools potentially imperils the market integrity and stability. The effectiveness of SEBI-introduced surveillance frameworks, including the graded surveillance measure and the enhanced surveillance measures, does not augur confidence. 

Further, the Surveillance Adequacy Index and the Index for Surveillance Effectiveness were framed to assess MIIs’ surveillance effectiveness. However, a spate of abusive trading instances frequently hogging the limelight indicates the need for better monitored markets.

What makes it worrisome is that the regulator is often found running behind the curve to detect manipulative market activities until incidental media reports, complaints, rating downgrades, or disclosures reveal abusive practices. Despite SEBI conducting suo moto investigations, including search and seizure operations, the slow pace of regulatory actions often leads to the perception of less-credible enforcement. 

Per SEBI reports, it took up 335 and 393 investigation cases related to insider trading, market manipulation, and price rigging in 2023-2024 and 2024-2025, respectively. However, it only completed investigations in 187 and 297 cases, signifying an accumulation of pending cases. 

Holistic Overhaul
A February 2025 IOSCO report titled ‘Thematic Review on Technological Challenges to Effective Market Surveillance’ underlines concern over regulators’ organisational and technical capabilities, highlighting their inability in integrated analysis of order and trade data from multiple venues. Further, it emphasises the need for cross-border surveillance capabilities. The sizable trading by foreign portfolio investors through offshore centres and instances of their alleged violations necessitate closer scrutiny of this aspect. 

The SEBI Market Surveillance System, built under the FIRE II initiative around 2005, has undergone many upgrades and transformations. SEBI highlights enhancing its market monitoring, detection, and investigation capabilities using AI/ML, NLP, network visualisation, and graph database-based analytics developed in its in-house lab, along with incorporating unconventional data sources and advanced technologies. Beyond experimentation of novelty use cases, these require enterprise-grade scaling and integration with digital workflows to achieve desired supervision quality and impactful results. 

Futuristic Surveillance 
The changing market structure requires enhanced surveillance abilities for real-time analysis and detection of manipulative market activities in correlated products and cross-segment frames. SEBI should adopt a zero-trust approach, similar to cyber-resilience, focusing on a preventive bias and predictive evaluations of potential violations. This approach demands focused investment in organisational capabilities, allocation of supervisory manpower and resources, and training in evolving surveillance technology systems and tools.

The key challenge in surveillance systems is handling voluminous false positives to reduce their noise. Enhancing alert precision with confidence scores can make a vital difference by reducing cumbersome reviews. It requires contextually adaptive pattern models to support dynamically calibrated rules and thresholds aligned to market and ecosystem factors. Assimilating historical trading and microstructural pattern insights enables spotting anomalies and deviations and validating the relevancy of suspicious outliers. 

The integration of AI/ML-driven pattern models can enhance real-time discovery of abusive trading patterns, adding predictive capabilities to surveillance with its adaptability to diverse scenarios and events. Automated review of severity-graded alerts, summarisation, case allocation, investigation, and evidence collection can be achieved through LLM and Agentic AI. Intelligent agents can orchestrate integrated workflows across stages, reducing manual reviews and interventions. Transforming scattered enterprise knowledge in paper-forms or scanned documents into machine-readable data becomes a critical requisite.

A data-driven regulator’s futuristic surveillance capabilities rely on an open and flexible data integration layer. Analysis of market and transaction data, aided with derived insights from non-traditional sources like chats, mails, blogs, news, disclosures, and public data, can help in real-time behavioural profiling and predicting the market abuse propensity of participants.

Data visualisation, replay, and reconstruction of trading activities for an entity across segments can uncover hidden abusive patterns. Beyond legal and regulatory boundaries, data sharing with other regulatory authorities like RBI, Insolvency and Bankruptcy Board of India, and Ministry of Corporate Affairs can create a more holistic approach, breaking siloed analysis of suspected entities across regulatory domains.

To remain ahead of the curve, SEBI should focus on addressing technical challenges and organisational resource gaps for scaling its present market surveillance capabilities. Prioritisation of investigations in big fish cases with timely and transparent enforcement actions can significantly enhance regulatory credibility and market confidence. India’s aspiration to become the world’s front-ranking market hinges on unwavering integrity and fairness in its market ecosystem.