Author : Mansi Tiwari from Navrachana University and Co-Author : Ansh Jain from Navrachana University
Abstract
The emergence of Artificial Intelligence (AI) technology has revolutionized the process of securities trading by allowing algorithmic trading platforms to scan news, social media posts, market movements, and political developments in near real-time and make trade decisions at lightning-fast speeds. Although AI-enabled algorithmic trading has made markets more efficient and liquid, it has also created serious issues such as market manipulation, algorithmic accountability, systemic risk, and investor protection.
In this context, this study will assess the phenomenon of AI-enabled algorithmic trading in India’s stock market, focusing on the sufficiency of existing legislation such as the SEBI Act 1992, SEBI (Fraudulent and Unfair Trade Practices) Regulations 2003, SEBI (Insider Trading) Regulations 2015, and SEBI’s Algorithmic Trading reforms 2025–2026.
It uses the term “AI-Driven Information Arbitrage” to describe how sophisticated algorithms leverage their speed and sentiments and alternative information to create market advantages. The paper highlights regulatory gaps concerning explainability of AI, governance of alternative information, liabilities for algorithms, and cross-border enforcement, offering recommendations to improve these areas.
Introduction
The incorporation of Artificial Intelligence (AI) in the financial markets has led to changes in how transactions in securities occur and risks and investments are managed. The traditional approach relied on human input, fundamental analysis, and slow information spread. In today’s scenario, however, trading algorithms using machine learning, NLP, and predictive analytics can process large amounts of information and perform trades in a matter of seconds without human intervention. This is the reason why the modern-day financial markets are no longer dominated by human traders but by machine-learning-based algorithms that learn from data and adapt accordingly.
The growth of India’s digital economy and FinTech sector has contributed significantly to this shift in the country. On one hand, the adoption of technologies such as UPI and digital banking and brokerage has enhanced access to financial markets, and on the other hand, it has resulted in the use of AI-based trading methods in India. There is a significant presence of algorithmic trading in India’s stock exchanges, especially in derivatives.
One case where AI is increasingly becoming dominant in shaping the outcome of market activities took place on 7th April 2025 after the “Liberation Day” tariff decisions made by former President of the US, Donald Trump. In response to the decision, the BSE Sensex went down by about 4,000 points, and the Nifty 50 index dropped under the 21,750 marks. In total, market capitalization worth ₹16 lakh crore was lost, while India VIX saw a rise by more than 56 percent. After further analysis of this case, it was established that a considerable number of trades was performed by AI algorithms, which analysed the decision, predicted its impact on the economy and triggered the formation of thousands of sell positions before the reaction of regular investors was registered.
The increasing use of AI-based trading systems has implications from a legal and regulatory perspective. Current algorithms employ the use of unconventional data inputs like financial news, social media, geopolitical events, satellite imaging, and economic data to make market predictions. Even though such advanced technology leads to increased market efficiency, liquidity, and price discovery, there are certain challenges associated with market manipulation, information inequality, algorithmic collusion, systemic instability, and the misuse of unregulated data inputs. Moreover, since many of these machine learning models are not transparent in their decision-making process, it is hard to justify why a specific trading decision was made.
The regulatory landscape in India has shown some signs of adaptation to these changes. Circular issued by SEBI on 4 February 2025 laid down the Algo-ID framework, strengthened responsibility for brokers, and imposed stricter supervision of algorithmic trading systems. It came into force from 1 April 2026. However, the current legal regime including the Securities and Exchange Board of India Act, 1992; SEBI (Prohibition of Fraudulent and Unfair Trade Practices) Regulations, 2003; Information Technology Act, 2000; and Digital Personal Data Protection Act 2023,was primarily meant to regulate actions undertaken by humans and traditional financial intermediaries. Thus, the current laws offer few solutions to challenges related to algorithmic decision-making, algorithmic responsibility, explainability, alternative data, and liability arising from disruption caused by machines.
This paper attempts to evaluate the rise of machine-driven algorithmic trading in India’s information-based financial markets and examine the adequacy of the existing legal framework in dealing with the new technological challenges. This paper argues that while India has been successful in introducing several regulations to govern the area of algorithmic trading, many important questions remain unanswered. It further explores how other nations have approached this problem and provides suggestions on possible reforms in India.
- AI-Driven Algorithmic Trading Process
The modern AI-based trading systems in India follow a multi-layered pipeline approach where news and data get converted into trade orders within a fraction of a second. Knowing the process is key to understanding its regulatory significance.
The Six-Step Trading Pipeline
Step 1 – Data Ingestion: Ingested live feeds include real-time tick data from NSE/BSE, Reuters/Bloomberg news wires, social media feeds, Reserve Bank of India circulars, and macroeconomic indicators across the globe.
Step 2 – NLP Sentiment Analysis Module: Transformers-based models (BERT and FINBERT) analyse headlines, social media content, and earnings calls. Named-entity recognition extracts companies, policies, geopolitical actors, and assigns numerical sentiment scores in milliseconds.
Step 3 – Signal Generation: Sentiment scores are aggregated with technical analysis metrics (RSI, MACD, VWAP) and order book statistics to generate directional BUY/SELL/HOLD signals weighted by confidence level.
Step 4 – Risk Management System: Stop loss logic, trade sizing, drawdown constraints, and circuit breakers prevent excessive trading losses. Simultaneously, Value-at-Risk (VAR) and Conditional VAR models compute portfolio risk constraints.
Step 5 – Execution Engine: A smart order routing system breaks down large trades between NSE and BSE to minimize trading impact. The latency goal is sub-milliseconds for high-frequency strategies utilizing co-location facilities at exchanges.
Step 6 – Feedback & Learning: Reinforcement Learning systems optimize model parameters via profit/loss feedback, thereby enabling automatic adaptation of machine learning trading systems to changing market regimes (bullish, bearish, or range-bound).
1.1 Practical Implementation of NLP Sentiment Trading
Probably the most important, and potentially illegal, element of AI trading today lies in its NLP capabilities. The classification of a tweet by Trump on Indian exports being “strongly negative,” for instance, might generate a −0.87 sentiment score, leading to trades in favor of selling in Nifty futures, Indian Information Technology stocks, and other exporting stocks within 3-5 milliseconds. Such systems respond not just to events that are certain to happen but trade on sentiments based on social media postings, satellite information, ship data, and even web scrapings of alternative information. The “alternative data economy” is giving rise to a two-tiered market where only those with computing power gain proxies for non-public information.
1.2 Categories of Algorithms Under SEBI’s 2026 Framework
The use of AI in algorithmic trading in India involves different approaches like HFT, market making, which benefits from the bid-ask spread, statistical arbitrage, which relies on price disparities, and NLP-based trading, which leverages the sentiments from news and social media platforms. Other examples are momentum-based algorithms and reinforcement learning programs, which learn by adapting themselves from market information. Though most of these approaches are permissible in the SEBI’s Algo-ID system, self-learning AI algorithms represent an emerging challenge due to their associated lack of guidelines and accountability mechanisms regarding AI ethics.
- Advantages of AI-Driven Algorithmic Trading
There are several positive economic and social factors arising out of the rise of algorithmic trading in India that need to be considered in any well-balanced analysis of this issue.
Market efficiency: Algorithmic trading reduces bid-offer spread, improves price discovery, and ensures that public information is quickly incorporated into asset pricing. In India, the AI Trading Platform Market is estimated to grow at a Compound Annual Growth Rate of 24.9 per cent from 2025 to 2030 and reach USD 2.3 billion.
Liquidity improvement: HFTs offer two-way continuous quotations in NSE and BSE, lowering the costs of transactions for both retail and institutional players by democratizing market access.
Precise trade execution: Smart Order Routing yields better fill prices as compared to manual trading due to its ability to minimize market impact cost for institutional portfolio managers.
Consistent risk management: AI trading incorporates risk controls such as position limits, drawdown control and correlation analysis at a pace and accuracy that human beings are unable to match, especially during times of turbulence.
- Disadvantages, Risks and Systemic Concerns
The same swiftness and scale that make AI-powered trading more efficient also make it hazardous for India’s financial markets, which have seen instances where such trading strategies have resulted in increasing market volatility, loss of investor confidence, and, most importantly, issues of equity.
Flash Crash: Multiple algorithms using identical NLP models are likely to engage in trading in the same manner when triggered by identical news in a matter of milliseconds, causing price spikes. The flash crash of the Trump tariffs of April 2025 in India saw ₹16 lakh crore wiped out as a result of simultaneous selling decisions made by multiple AI algorithms.
Informational Advantage Inequality: Institutions that are privy to better data feeds, co-located servers with NSE, and NLP engine superiority will continue to front-run small investors with their delayed access to information, a problem only partially rectified by current SEBI regulations.
Collusion Risk Without Agreement: Research in the field indicates that AI algorithms can learn to collude in their pricing behaviour through reinforcement learning despite the lack of any coordination.
Issues of Opaqueness and Black Box Risk:
Deep learning models are, by definition, opaque technologies. Neither the regulator nor the compliance departments of the deploying firm itself will have full visibility into why a certain trade decision was taken, thus undermining the possibility of accountability under existing regulations.
Exploitation of Fake News:
Sentiment algorithms cannot differentiate between credible news and fake news. Creating fabricated stories on social media platforms for the purpose of triggering the sentiment algorithms constitutes an innovative means of market manipulation not envisioned in the existing PFUTP Regulations.
Regulatory Arbitrage Between Jurisdictions:
Firms using offshore servers (Singapore/Cayman Islands) to run their algorithms while trading in Indian markets take advantage of regulatory arbitrage in SEBI’s enforcement jurisdiction.
3.2 The “Liberation Day” Case Study – April 7, 2025
The crash that took place on April 7, 2025, deserves special attention as a classic case of news trading driven by AI technology. President Trump’s declaration of 27 per cent tariff on all Indian exports (which, according to his “Make America Wealthy Again” speech, is intended as revenge against Indian restrictions on trade) was issued through White House official press releases and social media platforms at the same time.
Trading systems utilizing machine learning techniques and natural language processing engines instantly reacted to the news, and classified it as strongly bearish in relation to Indian exporting industries (information technology, pharmaceuticals, textiles), and for all risk-on portfolios in general. This resulted in a massive sell off from AI trading programs, causing the Sensex to plunge around 4,000 points and India VIX to soar 56 per cent above the threshold.
- Data Protection and Privacy in the FinTech Context
The passage of the Digital Personal Data Protection Act, 2023 (DPDPA) marks an evolutionary step towards a more secure FinTech future. The DPDPA creates a regime based on the concept of “informed consent” for the processing of “digital personal data” with detailed rules for the “significant data fiduciaries” (SDF), who shall undertake additional responsibilities like data protection impact assessment, designation of data protection officer, and algorithmic audit.
Among key compliance requirements of the DPDPA for FinTech organizations include: obtaining free, specific, and unambiguous informed consent for data processing; limitation of data processing to what is necessary for the intended purpose; retention of data for only the duration required; and maintenance of accuracy in the personal data processed. The DPDPA also provides data principal (i.e., individual) access, right to correction, erasure, and grievance redressal rights, all of which have significant operational implications for credit bureaus, digital lenders, and insurance aggregators.
In terms of how the DPDPA dovetails with the sector-specific rules on data localization, the legal question becomes quite complex.
Privacy is a constitutionally recognized fundamental right
within Article 21 of Justice K.S. Puttaswamy (Retd.) vs. Union of India (2017) 10 SCC 1. The constitutional right to privacy lays the groundwork for data protection within financial services. Cases relating to the data management policies of FinTechs can now be fought not only using the DPDPA’s administrative processes but also using writ petitions under the Constitution.
- Regulatory Loop-holes, Manipulation Opportunities & Legal Gaps
Even as SEBI strengthens its regulatory stance, there still exist major loop-holes that are cleverly used by smart market players, thereby creating systemic risks for India’s capital market environment.
5.1 The Gap Created by Fake News Manipulation
Present laws do not allow the spread of any sort of information in order to distort the market. But in cases where automated systems act upon false information put out on social media platforms – causing prices to move through the cascading effect of the algorithms – it becomes highly complex to identify the cause-and-effect relationship between the false news and the market action, thereby making detection difficult for SEBI.
5.2 Sentiment Data as Unregulated Inside Information
Sentiment feeds, satellite images of business establishments, credit card transactional data, and shipping documents serve as AI-driven trading systems’ material inputs even before their conventional release into the public domain. While the provisions of SEBI’s (Prohibition of Insider Trading) Regulations, 2015 are silent about alternative data as “unpublished price sensitive information,” this regulatory void is exploited by savvy participants who have an informational edge.
5.3 Overseas Hosting of Infrastructure to Evade Compliance with Algo-ID
While SEBI’s 2026 framework requires algorithm-hosting infrastructure be established in the brokers’ systems, firms having their algorithm-hosting strategies domiciled in Singapore and Cayman Islands and accessing India’s financial markets through authorised intermediary brokers can still escape the oversight of SEBI regulations. Enforcing the Algo-ID compliance requirements against offshore-domiciled algorithm strategies is quite difficult without bilateral regulatory cooperation agreements.
5.4 Absence of Liability for Algorithmic Flash Event Crashes
In case there is a flash event due to cascading failure caused by an algorithm, there is no framework in Indian civil or criminal laws to provide for compensation to the retail investors. Further, it is uncertain whether brokers face strict liability as implied under SEBI’s 2026 framework.
- Recommended Reforms to SEBI: Three-Pronged Approach
In consideration of the current regulatory framework and associated loopholes, the following proposals are made by this research to strengthen SEBI and India.
Proposed Pillar I: AI Ethical and Explainability Compliance System
Requirement of model documentation: any trading system using AI that exceeds certain trading volumes is required to provide model documentation explaining the reasoning process for the generation of signals.
Logging requirements of the algorithmic trading process: the logging of all inputs and intermediate results of the algorithm to allow for the auditing of any manipulation claims post facto.
Banning completely opaque black box trading models for retail trading strategies: models should be at least partially transparent based on a GDPR-compliant right to explanation principle.
SEBI-approved AI conduct assessments conducted annually by independent third party auditors, similar to financial audits.
III. Circuit Intelligence Framework and Systemic Risk Protocol
Real-time coordinated algorithm detection: NSE and BSE to establish AI-based monitoring system able to detect pattern recognition of coordinated algorithmic behavior suggestive of cascade risk.
News-event circuit breakers: intelligent, hierarchical suspension of trading following the classification of a geopolitical news event, using natural language processing, that could lead to an algorithmic cascade.
Cross-border regulation: SEBI to reach agreement on data and law enforcement cooperation with the SEC, FCA, and MAS regarding offshore algorithmic traders.
Strict liability scheme for brokers of flash events beyond pre-defined volatility limits with required compensation programs for retail investors.
6.1 Comparative International Models
The international community has been much stricter regarding the regulation of AI in stock trading. For example, the EU prioritizes the registration and explainability of algorithms, the US is more concerned about market surveillance and circuit breakers, the UK implements individual accountability rules, and Singapore incorporates risk control mechanisms with regulatory sandboxes. While the SEBI’s Algo-ID guidelines of 2026 provide a solid base for regulating algorithmic trading, India requires more regulations concerning AI transparency, alternate data management, algorithmic auditing, and system risks.
- Conclusion
This paper concludes that India is currently facing one of its most important periods in terms of regulating algorithmic trading by means of AI. Recent market events have proven that there is still a lack of effective measures to address the issue as well as the fact that news-related AI can cause great turbulence within the market. While SEBI’s 2026 Algo-ID guidelines should be praised, other measures need to be introduced as well. It goes without saying that these efforts could ensure fair trading on Indian capital markets.
Reference
- SEBI (Prohibition of Insider Trading) Regulations, 2015.
- Securities Contracts (Regulation) Act, 1956.
- Information Technology Act, 2000.
- Competition Act, 2002.