Kunj Bhatia*
Introduction
In July 2025, the Securities and Exchange Board of India came face-to-face with a problem its existing regulations were never made to address. Jane Street, a U.S. based proprietary trading firm headquartered in New York, was accused of earning a staggering INR 735 crores worth of returns in a single day, allegedly with the help of market abuse, high-frequency trading algorithms (HFTs), and sophisticated algorithm-driven strategies. While SEBI released an interim order dated 3 July 2025, aiming to recover almost 4,900 crore rupees, prosecution remains doubtful. Meanwhile, for the first time in history, it is found that algorithmic trade overtook manual trading in the National Stock Exchange (NSE) cash market in 2024, which captured a record of 53% of trades, rising from just 14% in 2010.
At the center of this storm is the Indian trade market and its retail investors, who are victims of regulatory gaps the current set of regulations were never meant to fill. This essay argues that SEBI’s primary instruments for combating market manipulation, SEBI (Prohibition of Insider Trading) Regulations, 2015 (PIT) and SEBI (Prohibition of Fraudulent and Unfair Trade Practices relating to Securities Markets) Regulations, 2003 (PFUTP), are archaic and adverse to market participants’ interests, thus raising a need for reform and the introduction of a mechanism meant to deal with algorithmic trading.
Two Regulations, One Problem: Why PIT and PFUTP Cannot Address Algorithmic Trading
Drafted in 2015, the PIT regulations center substantially on preventing information asymmetry and protecting ordinary citizens. Under Regulation 2(1)(g) of the PIT, insider trading is defined as trading conducted by a “connected person” or anyone with access to Unpublished Price Sensitive Information (UPSI). As per Regulation 2(1)(n), UPSI refers to any non-public information about a company or its securities that could significantly impact their market price once it becomes publicly available. Unlike traditional insiders, however, AI does not directly rely on finance leaks or UPSI but leverages alternative data, such as satellite imagery, social media sentiment, geolocation trends, or supply chain indicators, to produce predictive insights. In the Indian context, AI-driven tools could analyse railway freight data or social media chatter around IPO-bound companies to project revenues ahead of formal announcements. When such predictions mirror the effect of UPSI, a critical question arises: Does this constitute insider trading or a sophisticated extension of legitimate analytics?
In addition, a widely accepted interpretation of PIT shifts liability onto human insiders and their behavior, which leads to the assumption that such acts would be identifiable and traceable to an individual. An algorithm does not “intend” in the way humans do, so can it be an “insider”? How then should liability be assigned if machines can replicate or even outperform human insiders? When an independent autonomous algorithm executes a trade based on data patterns that reveal UPSI, who is liable? The programmer or the organization that developed the algorithm, the trader who deployed it, the company providing the data feeds, or the algorithm itself as a quasi-legal person?
A dimension of algorithmic market manipulation emerges even in the absence of UPSI. With the rise of HFT, milliseconds of advantage can give groups unfair advantage over common investors and decide profitability. They capitalize on structural market inefficiencies through strategies like quote stuffing, where algorithms flood the market with fake orders to slow down competitors, often placing millions of orders per second or latency arbitrage, where the algorithm takes advantage of the price discrepancy of the same stock in two different markets due to just a few microseconds of delay in relaying of information between the two exchange data centres. This widely used controversial strategy came under major scrutiny in the Indian legal dimension during the NSE co-location scam when a whistleblower in 2015 sent a series of detailed letters directly to the market regulator SEBI and financial journalists, blowing the lid off how select algorithmic trading brokers were getting unfair early access to the NSE server and using HFT to earn crores of profit unethically.
While algorithmic malpractices can distort the market just as much as the practices covered under PIT and PFUTP, they fall under a regulatory grey zone simply because they take advantage of regulatory and technological loopholes rather than human misconduct.
Data, Intent and Opacity: Beyond India’s Regulatory Borders
To say that India’s anti-algorithmic manipulation regulations are toothless and completely inefficient would be a poor assessment of local capabilities. Patterns such as those observed in the Jane Street case are already flagged under SEBI’s Surveillance Manual (2021). Furthermore, two circulars, namely, the API-based Algo Framework (2022) and subsequently, SEBI’s Algorithmic Trading Circular (2025), both of whose provisions require all algorithmic strategies used in Indian markets to be pre-approved, audited, and lastly, are deployed with real-time risk controls left to the SEBI.
However, the inability of such regulations to answer who bears liability when an algorithm leverages UPSI without human intent, whether alternative data analytics constitute manipulation, or how Agentic AI systems should be governed when their creators cannot fully anticipate their behavior, exposes a critical gap in the regulatory framework. The Indian courts have been silent so far on these matters, and this raises the need to seek other comparative perspectives and review how the system can benefit from them.
The US Securities and Exchange Commission (SEC) recognizes the Mosaic Theory, which permits sophisticated funds to collect public, non-material, or alternative data points to piece together a valuable investment thesis. However, a critical boundary is drawn to distinguish between public and personal data. For example, if a hedge fund buys credit card transaction data or mobile location logs that contain Personally Identifiable Information (PII) or data obtained via a hacked server or a breached non-disclosure agreement (NDA), it instantly triggers enforcement actions under Rule 10b-5.
A second, equally pressing gap lies not in what data algorithms use, but in when their trades occur. Indian exchanges record order timestamps in milliseconds (one-thousandth of a second), while HFT logic, loops, while market manipulation operate in microseconds (one-millionth of a second). Without mandating a standard time protocol across all market participants, it is physically impossible for SEBI to reconstruct the precise sequence of events during high-speed flash crashes or circular trading loops like the European Securities and Markets Authority (ESMA) does under the Regulatory Technical Standard 25 (RTS 25) of the Markets in Financial Instruments Directive II (MiFID II).
A commonly used means used to bypass SEBI PIT rules is claiming an absence of “intent to trade based on that UPSI”, a requisite to prove any kind of insider trading or malpractice, mirroring the necessity of mens rea in criminal law. The perpetrators shift the blame onto the programmer codes, while the coders claim that they had control only over the means and not the end. In a traditional fraud case, a prosecutor must prove specific intent, but with high-frequency algorithmic systems, such intent does not exist. Therefore, the US Commodity Futures Trading Commission (CFTC) and the Securities and Exchange Commission (SEC) use Reckless Disregard theory, which means they evaluate the intent of the architecture, not the intent of an individual execution. A simple question is asked: Would an objective, competent software developer know that writing this specific code would inevitably distort the market? And based on the answer, the authorities move forward with the prosecution.
A coexisting global issue is the “Black Box” problem, where, while the system’s data inputs (such as order book depth or corporate feeds) and outputs (executed trades) are visible, the internal mathematical logic and neural decision-making path remain entirely hidden from investors, compliance officers, and regulators. The term XAI—Explainable AI—has thus gained traction in this field, with leading regulatory bodies all around the world leaning towards tools like SHAP (Shapley Additive exPlanations), whose frameworks mathematically break down the exact weight each data variable contributed to a final trading execution. Alongside these tools, EU’s MiFID II rules force institutions to run black boxes through a sandbox, which involves a simulated environment that mirrors a real-life stock exchange (like the NSE or BSE) but operates with zero real-world financial risk.
Conclusion
SEBI’s PIT and PFUTP regulations, drafted in an era when the securities market was still filled with human characters, were designed to punish mens rea and information asymmetry between human actors and not interrogate algorithms that anticipate market insights from satellite parking lots or track down liability through autonomous codes that execute trades in microseconds across jurisdictions.
India need not reinvent the ecosystem, and the comparative experience offers a roadmap. SEBI must move beyond ex post enforcement toward ex ante structural safeguards, which remain premised on a model of human-permitted and audited automation, not yet accounting for Agentic AI systems that learn, adapt, and execute in ways their creators neither fully comprehend nor anticipate. For such systems, pre-approval and real-time risk controls, along with a regulatory sandbox for emergent behavior, are necessary, where the decision trees are mapped to detect manipulation patterns in silico rather than in situ.
India’s insider trading regime was never built to prohibit wrongdoing by machines, but to punish the choices of people. Closing that gap will require SEBI to stop treating its rules as a fixed list of banned conduct and start treating them as a living system, one capable of keeping pace with actors that trade, learn, and cover their tracks faster than any human regulator possibly could.
* The author is a second-year Student of B.A. LL.B. (Hons.) at National Law University, Jodhpur.
This blog reflects the personal views of the author and does not necessarily represent the views of The Policy Chronicle.