(Anirudha Bhat and Devansh Shrivastava are Fifth-year students at the National Law School of India University)

India’s securities markets are increasingly being shaped by algorithms. Today, a substantial share of orders on Indian stock exchanges originates from automated trading systems capable of executing trades in milliseconds. While regulators have spent years developing safeguards against market manipulation and conventional insider trading, a more difficult question is beginning to emerge: what happens when an algorithm trades on an informational advantage that originated from unpublished price-sensitive information (UPSI)?

The Securities and Exchange Board of India (Prohibition of Insider Trading) Regulations, 2015 (PIT Regulations) were designed around a relatively simple understanding of insider trading. An individual obtains confidential information, decides to trade on the basis of that information, and profits before the market becomes aware of it. The PIT Regulations, therefore, focus on three concepts: (1) possession of UPSI, (2) communication of UPSI, and (3) trading while in possession of UPSI.

Algorithmic trading complicates this framework. Information can now be embedded into trading systems long before any trade is executed. In some cases, sophisticated systems may generate insights functionally equivalent to insider information without any identifiable act of human misuse. These developments raise difficult questions about attribution, liability, and regulatory design.

This blog argues that India’s human-centric insider trading framework is ill-equipped to address algorithmic insider trading and proposes reforms to close this gap.

The Human Assumptions Behind Insider Trading Law

The PIT Regulations rest upon three implicit assumptions. First, they assume the existence of a human actor who possesses information and makes a conscious decision to trade. Second, they assume a discrete informational event. UPSI is generally conceptualized as information that comes into existence at a specific point in time, is possessed by identifiable individuals, and subsequently influences trading decisions. Third, they assume a traceable causal connection between possession of information and execution of a trade. Regulators investigate whether a person who possessed UPSI later traded in securities connected to that information.

These assumptions worked reasonably well in traditional markets. However, algorithmic trading challenges all three simultaneously. Modern trading systems execute decisions through pre-programmed rules, statistical models, and increasingly, machine learning techniques. Human involvement may occur only during the design stage, while execution is entirely automated. Consequently, the moment at which the informational advantage is embedded into the system may be separated from the actual trade by days, weeks, or even months. The result is a regulatory mismatch between legal doctrine and technological reality.

The Problem of Parameter-Setting

The most immediate challenge arises from what may be called parameter-setting insider trading. Every algorithm operates according to predefined parameters. These include price thresholds, trading triggers, volume limits, timing windows, and risk-management rules. The configuration of these parameters determines how the system behaves under different market conditions.

Imagine a senior executive learns that the company is about to announce unexpectedly poor quarterly earnings. Instead of personally selling shares, the executive influences the design of an algorithmic strategy. The system is configured to exit positions once certain market indicators appear or once the stock reaches a predetermined level.

When the algorithm eventually executes the trade, there may be no obvious evidence of insider trading. The trading activity appears routine. Yet the informational advantage was embedded in the system from the beginning. The legally significant decision was not the trade itself but the configuration of the algorithm.

This creates an enforcement problem. Existing regulatory systems primarily observe trading outcomes. They record what an algorithm did. They rarely capture why specific parameters were selected or what information was available to the person designing the strategy. Consequently, the informational advantage may disappear into the architecture of the system itself.

Why Existing Law Struggles

The PIT Regulations prohibit trading while in possession of UPSI and prohibit the communication or procurement of UPSI except for legitimate purposes. At first glance, these provisions appear broad enough to cover many technological developments. However, on a closer examination, there are some important limitations.

The concept of possession remains closely tied to human actors. Regulation 4 asks whether a person traded while in possession of UPSI. The provision becomes more difficult to apply when the relevant decision was made during algorithm design rather than at the moment of execution.

Similarly, the prohibition on communication of UPSI presupposes identifiable human conduct. It is relatively straightforward to prove that an insider shared confidential information with another individual. It is far more difficult to demonstrate that confidential information was encoded into a trading model through a series of design choices.

The challenge becomes even greater when machine learning systems are involved. Advanced Algorithms often operate as “black boxes.” They identify patterns and generate predictions that may not be fully explainable even to their developers. A regulator may observe that a system consistently trades ahead of significant market events, yet proving that the system relied upon UPSI rather than sophisticated data analysis becomes extraordinarily difficult. The law’s focus on human intention and possession, therefore, sits uneasily with technologies that separate decision-making from execution.

The Attribution Gap

Perhaps the most significant regulatory challenge is attribution. Insider trading law ultimately seeks to identify a responsible actor. Liability must be attributed to someone. In traditional cases, this process is relatively straightforward. An executive receives confidential information and executes a trade. The same person possesses the information and undertakes the relevant market activity.

Algorithmic systems disrupt this link. The developer who designed the algorithm may never execute a trade. The trader who deploys the algorithm may never possess UPSI. The system itself cannot be treated as a legal person. Meanwhile, execution occurs automatically through software. This creates what may be described as an attribution gap. The informational advantage exists, the trade occurs, but the connection between the two becomes difficult to establish.

The problem becomes particularly acute where algorithms operate autonomously over long periods. A person may possess UPSI while configuring a model in January, yet the resulting trades may occur months later. By that point, the information may have become public, or the individual responsible for the design may no longer be associated with the firm. Current regulatory frameworks are poorly equipped to address such temporal separation.

The Limits of Existing Enforcement

SEBI has developed increasingly sophisticated frameworks governing algorithmic trading. These frameworks focus on market integrity concerns such as spoofing, layering, and other forms of manipulative conduct. However, insider trading presents a different challenge.

Manipulation is often visible through patterns of conduct. Regulators can identify unusual order placement, cancellation behaviour, or trading sequences. Insider trading through algorithmic systems is often invisible. The algorithm may behave perfectly normally. The informational advantage is hidden not in execution but in design.

This distinction matters because surveillance tools built to detect manipulation may not identify insider trading risks arising from algorithmic configuration. As algorithms become more sophisticated, the gap between surveillance capabilities and actual informational misuse may continue to widen.

Rethinking Insider Trading Regulation

Addressing these challenges requires moving beyond a purely human-centric model of regulation.

One important reform would be to expand the concept of trading itself. Current frameworks focus primarily on the execution of transactions. Yet in algorithmic environments, the economically significant decision often occurs when trading logic is created or modified.

Regulators should therefore recognize parameter-setting and strategic model configuration within the definition of trading under regulation 2(1)(l).

A second reform involves creating explicit parameter-setting liability for parameter setting. Individuals who design algorithmic strategies should bear responsibility where those strategies are configured using UPSI. Such an approach would establish a clearer legal connection between the “controlling mind” and the subsequent conduct of the algorithm.

Third, firms should be required to maintain comprehensive algorithmic audit trails. Every significant modification to trading logic should be recorded, together with information regarding who authorized the change and what data sources informed the decision. These records would provide regulators with a much clearer picture of how trading systems evolve over time.

Finally, greater emphasis must be placed on explainability and accountability in Algo-driven financial systems. As algorithmic trading becomes increasingly influential in market activity, regulators will need mechanisms capable of reconstructing the decision pathways that lead to particular trades. Without such safeguards, enforcement risks becoming increasingly ineffective.

Conclusion

India’s PIT Regulations remain anchored in assumptions about possession, intention, and execution that emerged in an era of human traders. Algorithmic systems challenge each of these assumptions by separating information from execution, embedding advantages within system design, and obscuring the causal relationship between knowledge and trading activity.

This does not mean that existing insider trading law is obsolete. However, it does suggest that the regulatory framework requires adaptation. Concepts such as parameter-setting liability, algorithmic audit trails, and explainability may become essential components of future market regulation.

As artificial intelligence becomes increasingly integrated into financial markets, the central challenge for SEBI will not simply be preventing insider trading. It will be ensuring that accountability survives the transition from human traders to autonomous systems.