By Taru Jhunjhunwala and Saloni Maheswari

INTRODUCTION

To address and regulate the growing use of artificial intelligence within the legal fraternity, the Supreme Court recently issued the Draft Regulations for Use of Artificial Intelligence in Courts, 2026 [hereinafter ‘Draft Regulations’] inviting comments from the general public. This is a much appreciated step taken by the Court especially in light of recent incidents involving AI hallucinations like the one wherein a lower court in Andhra Pradesh relied upon fake AI-generated judgements to adjudicate a property dispute. It was, therefore, necessary for the SC to set boundaries and provide guidelines governing responsible AI usage by litigants and more importantly by judges and other officers of the Court.

Chapter XI provides for grievance redressal, with Regulation 52 addressing issues arising from prohibited AI use and Regulation 53 preserving remedies available under other laws. While these provisions appear comprehensive at first glance, they fail to adequately address grievances arising from the permissible use of AI under Regulation 19.

SCOPE OF GRIEVANCE REDRESSAL MECHANISM

Regulation 52 establishes a grievance redressal mechanism for harms arising from the prohibited use of AI under Regulation 20. However, in a world where every AI tool comes with a disclaimer that AI might make mistakes, it would be wrong to presume that permitted AI usage cannot cause harm.

The question that comes to mind now is what happens when the harm is caused to any person due to the lawfully permitted usage of AI as given under Regulation 19? The answer to this lies under Regulation 53, where it is provided that any person affected by the permissible usage of AI within these guidelines has the option to approach the court under any other applicable law. Prim facie Regulation 53 solves the issue about the risks which come along with the usage of AI. However, for this grievance redressal mechanism to be effective, there needs to be effective remedies available under other laws. Before delving into the remedies, it is important to look at Regulation 19 to see what kind of and what extent of harm can be caused by the permissible usage of AI.

When Authorised AI Causes Harm

Regulation 19 permits AI usage in two broad categories. Sub-regulations 19(b) and 19(c) require human verification before AI-generated outputs are acted upon. Consequently, the likelihood of harm is lower and liability can generally be attributed to the supervising individual. In contrast, sub-regulations 19(a), 19(d), and 19(e) permit AI use without mandatory human oversight. Although these provisions relate primarily to administrative and clerical functions, the potential harm can be significant. For instance, Regulation 19(a) allows AI to be used for case management and cause list preparation. The concern that comes with this usage of AI is that AI may incorrectly flag a valid filing as defective or might fail to identify the actual defects. To illustrate the intensity of the harm, let us assume that there needs to be a hearing for an urgent bail application, but it is incorrectly marked as incomplete and thereby is not processed promptly.

Another concern is that training data for AI may contain certain historical biases, as a result of which certain categories of litigants might receive less favourable administrative treatment. There are plenty of reports/surveys which highlight that AI generated answers contain some pre-existing stereotypes. Moreover, if we look at Regulation 19(d), it provides AI usage for legal research, document verification, etc without any human supervision. This step seems progressive, however we should not ignore the fact that the risk of AI hallucinating cases is common. It might also be the case that AI may systematically prioritise certain types of cases, or viewpoints.

If these biases or stereotypes are reflected in courtroom management, where people actually come to seek redress from these stereotypes only, it will shake the public trust in the judiciary. Hence, it is very necessary to provide a dedicated grievance redressal mechanism for such unwanted harms.

Remedies in a Vacuum

The Digital Personal Data Protection Act (DPDP), 2023 [hereinafter ‘DPDP Act’] addresses harms arising from the processing of personal data. For instance, AI systems used for case listing may analyse information such as property records or personal details to determine the nature of a case or the appropriate Bench. Any harm resulting from the misuse or incorrect processing of such personal data may be addressed under the DPDP Act. However, many AI-assisted judicial functions, such as case status management, case summarisation, or legal research, involve little or no personal data beyond information already available in the public domain. In such cases, the DPDP Act offers limited recourse.

Hence, for a progressive AI policy to be effective, it must be supported by a robust grievance redressal mechanism, drawing on best practices from other jurisdictions.

GLOBAL BEST PRACTICES

General Data Protection Regulation

Under Article 15 of General Data Protection Regulation [hereinafter ‘GDPR’] of the European Union, if a process involves automated decision making i.e. a decision arrived at by using AI, then the data subject i.e. the person whose data was used, is entitled to have meaningful information about the logic involved in the decision-making system. The European Court of Justice recently in CK v Magistrat der Stadt Wien clarified the scope of “meaningful information” under Article 15(1)(h) and held that data controllers are required to provide concise, transparent, intelligible and easily accessible explanations of “the procedures and principles actually applied”.  Thus, the purpose of this right is not merely to inform individuals but also to make them understand the decision and challenge it wherever necessary.

However, GDPR is European law and is not applicable on Indian soil. India’s national data regime i.e. the DPDP Act does not contain any provision equivalent to right to explanation provided under the GDPR.

Data Protection and Data Privacy Act

Section 11 of the DPDP Act that provides individuals the right to access information about their data is silent on the aspect of automated decision-making process. Furthermore, the DPDP Act also lacks any provision similar to Article 22 of GDPR which deals with automated decision making and profiling. This results in a major regulatory gap, leading individuals virtually without any remedy if AI usage for example leads to biased cause lists giving priority to a specific set of individuals on the basis of caste, creed etc.

EU Artificial Intelligence Act

Additionally, Article 86 of the EU Artificial Intelligence Act [hereinafter ‘EU AI Act’], provides that individuals affected by decisions involving highrisk AI systems are entitled to receive a clear explanation of the AI system’s role in the decision making process and the parameters it considered to arrive at that decision. Moreover, the EU AI Act classifies such judicial administrative systems as high risk, thus, attracting Article 86 and subjecting the judicial system to strict transparency requirements.

UNESCO Guidelines

According to the UNESCO Guidelines for the Use of AI Systems in Courts and Tribunals [hereinafter ‘UNESCO Guidelines’], judicial systems using AI should endeavour to ensure accountability and contestability by “informing and explaining how and why certain AI tools were adopted by the judiciary and the results achieved with their deployment”. Thus, the UNESCO Guidelines create two obligations, first, to inform parties when AI systems are being used and second, to provide information when necessary to determine liability. The existing draft regulations contain neither of the two, i.e., litigants would neither know for example whether AI was used to process their filings nor would they have a right to seek an explanation if an AI-induced error takes place. While these guidelines are mere recommendations and not binding on member states, India endorsed the UNESCO Recommendations on the Ethics of Artificial Intelligence in 2021 which also emphasise upon accountability and transparency.

The GDPR, the EU AI Act and the UNESCO guidelines point towards a higher standard of transparency and an individual’s right to be able to understand how AI had an impact on their rights and liabilities. Since the DPDP Act does not provide any remedies, the Draft Regulations should fill this gap and ensure that there is accountability in the true sense for AI usage in courts.

GIVING MEANING TO THE RIGHT TO EXPLANATION

It is important to clarify how such a right to explanation should operate within the framework of the Draft Regulations. Litigants should be notified whenever an AI system plays a material role in an administrative decision making which materially affects their case.

Any person adversely affected by such an outcome should be entitled to seek an explanation within a prescribed period. Upon receiving such a request, the Court should provide sufficient information to enable the affected party to understand the basis of the decision. This could be done by providing a detailed statement containing (i) the purpose for which the AI system was used; (ii) the inputs given to the model; and (iii) the reasons behind the administrative decision.

The Court should also be required to maintain records of such AI-assisted acts which would be helpful in determining whether an error originated from the AI system or the data provided to it etc. Additionally, where it is shown that an adverse effect was caused, the litigant should have access to a review mechanism consisting of judicial officers and persons with technical expertise.

CONCLUSION

The Draft Regulations are a timely attempt to guide the use of AI within the judicial system. However, the effectiveness of any regulatory framework ultimately depends not only on what it permits or prohibits, but also on the remedies it offers when things go wrong. The present framework addresses harms arising from prohibited uses of AI, but remains largely silent on the consequences of errors resulting from authorised uses. Global developments show a growing emphasis on transparency, explainability, and accountability in AI-assisted decision-making. Incorporating similar safeguards within the Regulations would strengthen public confidence in the judiciary and ensure that individuals are not left without recourse when AI-driven processes affect their rights.

(Taru Jhunjhunwala is a fourth-year B.A. LL.B. (Hons.) student at National Law University, Jodhpur. Saloni Maheswari is a fourth-year student at NALSAR University of Law, Hyderabad. The authors may be contacted via mail at tarukumari.jhunjhunwala@nlujodhpur.ac.in and salonikothari125@nalsar.ac.in respectively.)