Abstract

Artificial Intelligence (AI) is rapidly moving from closed testing environments into real-world execution across several sectors. The incorporation of AI into judicial frameworks is now a reality. Unlike other sectors, judicial decision-making is bound to human reasoning, and unchecked AI deployment raises serious concerns about due process and the protection of fundamental rights. This article analyzes the judicial rationales behind the Supreme Court of India’s guidelines on AI and situates them within comparative global frameworks. It evaluates where India’s approach and guidelines stand against prevailing international standards. This analysis reveals both the strengths of India’s approach and the structural questions that remain open.

Artificial Intelligence (AI) is rapidly moving from closed testing environments into real-world execution across several sectors. The incorporation of AI into judicial frameworks is now a reality. Unlike other sectors, judicial decision-making is bound to human reasoning, and unchecked AI deployment raises concerns about due process and the protection of fundamental rights. This article analyzes judicial rationales behind the Supreme Court’s guidelines on AI and situates them within comparative global frameworks. It evaluates where India’s approach and guidelines stand against the prevailing international standards. This analysis reveals both the strengths of India’s approach and the structural questions that remain open.

Tracing the Institutional Path: India’s Judicial AI Framework

The Supreme Court’s engagement with artificial intelligence in the court process has been gradual and deliberate. In 2021, the Supreme Court launched India’s first AI-driven research portal – SUPACE (Supreme Court Portal for Assistance in Court’s Efficiency). It was the first step taken by the Supreme Court to leverage AIinthe judicial system and to build a future-ready judiciary. Developed in cooperation between the Supreme Court of India and IIT Madras, it intends to develop a module to understand the factual matrix of cases with an intelligent search of precedents, apart from identifying the relevant cases.

The recent guidelines represent a natural progression of this process. Recently, in June 2026, the Supreme Court of India released the Draft Regulations for Use of Artificial Intelligence in Courts, 2026, inviting public comments until 20^th June. This is an institutional attempt by the Supreme Court to govern the deployment of AI across the entire judicial system.  The draft specifically emphasizes that the use of AI in the court process should “remain strictly subservient to human oversight and judicial authority”. It treats AI as an instrument of assistance rather than a replacement for humans in the process, extending constitutional continuity into the draft, and maintaining a balance between constitutional safeguards and technological advancement. It cautions against the risks of widening the digital divide and the exclusion of linguistic and cultural minorities.  Further, it does not treat the judicial process as a monolith; rather, it disaggregates it into administrative and adjudicatory functions, applying a fundamentally different governance standard to each. It further institutionalizes continuous oversight and scrutiny of the functioning and development of AI in the judicial domain, creating a web of specialized committees and appropriate authority.

The framework is designed to be forward-looking and technological-neutral, even looking to govern future advancements in AI deployment. But this is an uncertain reasoning as neither the full impact of the disruption nor the future trajectory is fully known. Therefore, it can be best understood as a starting point rather than a definitive settlement.    

Inside the Draft: Structure, Features, Imperatives

The draft opens with articulating general definitions and principles related to AI deployment. It creates an unambiguous foundational vocabulary, which is crucial for uniform interpretation and application throughout the judicial process. It explains the technical AI terminologies like black box, Human-in-the-Loop, Large Language Model (LLM), Machine Learning (ML), and Algorithmic Decision making (ADM). Further, it addresses data privacy, cybersecurity, and data minimization as legally enforceable obligations. It situates them within the framework of the Information Technology Act, 2000, and the Digital Personal Data Protection Act, 2023. This legislative linkage signals that the Supreme Court intends to harmonize these guidelines with the existing statutory framework.

At the philosophical core of these guidelines, it is the principle of human primacy. The draft unequivocally mandates that AI must be used only in an assistive capacity and cannot replace judicial conscience. The rationale behind this is that the judicial order derives its legitimacy not just from its correctness but from the fact that a human being is accountable for the order. It clearly states that AI cannot replace judicial officers in determining questions of law, fact, or justice. The accountability for any decision taken with AI assistance continues to rest exclusively with the judicial officer. It clearly mentions that the outputs of an AI System, the opaqueness of a Black Box system, or the occurrence of hallucination cannot be taken as a ground for avoiding accountability for a palpably incorrect, illegal, or harmful decision. Further, the adoption and use of AI systems in courts shall be consistent with the provisions of the Constitution and any other statutory framework.

Further, the draft framework is rooted in the essence of Article 14: fairness and non-discrimination. It clearly stipulates that AI systems shall not be deployed in a manner that perpetuates, amplifies, or introduces bias and inequality on the grounds of race, religion, or caste, sex, gender, disability, language, economic status, or any other ground prohibited under the tenets of the constitution. The deployment of opaque or incapable explanations of AI systems in high-risk applications is restricted to protecting the rights of concerned persons. It ensures that the use of AI in the court process remains proportionate to the nature, complexity, and risk profile of the relevant task. To preserve data integrity, guidelines mandate that AI systems should be trained and operated based on data that is accurate, representative, lawfully obtained, and free from discriminatory bias. It states that data possessed by AI systems must be protected by robust, layered, and continuously updated organizational security measures.

The guidelines disaggregate the functioning of the judicial process into “adjudicatory and administrative functions”, resolving the central tension of judicial AI governance between the efficiency imperative and constitutional demand for human reasoning. It structurally specifies the permissible and prohibited use of AI systems. Essentially, the adjudicatory functions involving higher-risk applications attract a higher degree of human oversight than administrative tasks such as scheduling or document management. It also makes remedial provisions for violations of regulations.

Furthermore, the draft constitutes specific committees and organizational branches to oversee the development and deployment of AI systems. It establishes a permanent, full-time Apex body named Centre for Research and Excellence in Artificial Intelligence to regulate and promote innovation, integration, governance, and policy development on AI in the judiciary to assist the apex body in various aspects. 

This comprehensive framework of the Supreme Court’s draft, its constitutional grounding, and systemized oversight and basal precision raise broader questions. How does India’s approach to AI governance measure up against the prevalent global frameworks? A close examination of the EU, the US, and China reveals the aspects in which India’s approach is distinctive and, in some areas, ahead of existing frameworks.

Global Frameworks- EU, US and China Model^[i]

European Union Model

The European Union (EU) governs the use of AI in the court process through the 2024 AI Act. It is a comprehensive, binding legal framework that is based on a risk classification system. This classification is not just technical; it imposes a complex set of mandatory obligations that must be fulfilled before the deployment of AI systems. The most striking feature of these regulations is their emphasis on prevention rather than correction. It mandates the regular assessment and audits of deployed AI systems. Like India, the EU also places human oversight and accountability at the core of its structure. AI usage is subject to scrutiny to ensure the level of transparency and traceability of the functioning of high-risk AI systems. Further restrictions come from the EU’s strict data protection regime, the General Data Protection Regulation (GDPR), which imposes additional requirements for the lawful processing of personal data, data minimization, and purpose limitation.

The EU model ultimately reflects the policy of calibrated restraint, which neither prohibits nor embraces AI in the judicial process. It adopts the philosophy of technological neutrality, promoting AI as an assistive tool rather than a substitute for humans. It institutionalizes a system with extensive guardrails to preserve accountability, privacy, transparency, and public trust in the system.

United States Model

The United States reflects a very different picture from the EU’s extensive binding regulatory structure. It uses loose guidance, providing voluntary guidelines on identifying, assessing, and deploying AI systems in routine court proceedings, lacking a unified national framework governing the use of AI in the judicial process. Instead, it is divided across individual state legislatures, court guidelines, and executive orders, providing a patchwork of guidelines which varies across various jurisdictions.

The focus is more on regulating the use of AI by lawyers and litigants than AI systems themselves. For example, US states like Mississippi, Florida, California, and Nebraska have issued fines and suspensions for lawyers using unverified AI-generated fake cases in court proceedings.

The US Model represents a commitment to innovation and market-driven development. It centers more around correcting the flaws through litigation and market mechanisms rather than preemptive regulation. It reflects the deeply embedded American principle that regards excessive regulation as a menace to technological innovation and economic development.

China Model

China has been a pioneer in integrating AI into the judiciary. As far back as 2021, it issued the Chinese government white paper, which intends to “promote AI inclusion in applications involving collecting evidence, managing documents, and analyzing the cases”. China focuses on integrating AI at the local level, and it trains AI systems on case databases to create ‘similar cases’ systems that can predict outcomes based on precedents. At the national level, the Chinese Supreme Court developed a nationwide legal AI structure trained on massive, high-quality judicial data.

The most interesting aspect of this model is its experiment with specialized Internet courts. Situated in major urban centres, these courts handle disputes arising from online transactions, intellectual property, and digital commerce. These courts extensively use AI for analyzing case details, evidence verification, and preliminary dispute resolution. But unlike EU and Indian systems, the Chinese model is centered around the state itself. The concerning aspect is the absence of independent human oversight, transparency, and accountability. It is established on a foundational premise that the Chinese state is the sufficient and ultimate guardian of public interest in AI governance. 

The core idea of the Chinese model is to treat technological advancement and institutional efficiency as primary values, while governance safeguards are structured as subordinate to state objectives.

Where Indian Model Stands Compared to International Benchmarks

The Indian model is deeply aligned with the core values of the Indian Constitution. On an international scale, it shares its foundational principles with western democratic values. It shares the EU commitment to rights-based governance and focuses on transparency, human primacy, and data privacy. Like in the US, it places ultimate accountability on judges and litigants rather than on the AI systems themselves. It stands in contrast to the Chinese state-oriented and efficiency-driven framework.  It can easily be described as a mixture of the EU governance-heavy model and the US guidance-based approach.

A deep comparative analysis of these regulations against prevailing international benchmarks reveals several significant lacunae that warrant serious attention. The most notable gap in the draft is the concentration of regulatory authority within a single judicial institution. Under Regulations 18, 22, 33, 35 and 52: the appropriate authority is constituted as either an Apex body at the Supreme Court or an AI committee at the relevant High Court. This consolidates the responsibilities of system deployment, regulatory oversight, independent auditing, and dispute resolution within a single institutional framework, creating an inherent conflict of interest. Secondly, under Regulation 46(1), the approval gateway for private entities in the functioning of AI systems in court processes seems rigorous on its face, requiring comprehensive regulatory approval by the Appropriate Authority under Regulation 46(2). Yet, the draft provides no criteria, no minimum standards, no scoring methodology, and no timeline for this evaluation. The appropriate authority has complete discretion over the entire process with no provision for independent audits rendering it opaque. It potentially violates “Wednesbury reasonableness” standards and Article 14’s prohibition on arbitrariness, as articulated in landmark cases of E.P. Royappa and Maneka Gandhi.  Thirdly, under Regulation 46(6), private vendors are required to report data breaches to the appropriate authority, but there is no penalty regime for failure to fulfill this requirement. There is no obligation to notify the litigants and no mandatory reporting to the Data Protection Board established under the Digital Personal Data Protection Act, 2023. Further, under Regulation 46 (4), there is no protection for the affected litigant under the indemnity clause. Lastly, the most fundamental concern is that these regulations are framed by the Supreme Court under the court’s rule-making power under Article 145. This makes these regulations entirely procedural in character, which means that they govern court practice, not substantive rights or technology governance. Regulations under Article 145 do not carry an independent penalty regime for violators. It is not suitable for regulatory violations like data breaches, algorithmic bias incidents, or vendor non-compliance. It places the court in an unfavorable position of being simultaneously the regulator, the deployer, and the enforcer. This creates a structural conflict in the draft. For meaningful progress in AI governance in India, there is a need for a dedicated, legislative-backed effort.

Reforming the Draft: Three Lessons from the EU AI Act

Inspiration can be taken from the EU Artificial Intelligence Act 2024, which offers several concrete and implementable solutions for the Indian AI judicial framework. Firstly, to curb the concentration of authority within a single judicial institution, we can create a mechanism of regular assessment by independent bodies constituted by AI experts. These bodies must be structurally, financially, and functionally separated from institutions that deploy AI systems. India can establish a judicial regulatory authority that can be modelled analogously to the Securities and Exchange Board of India (SEBI) with supervisory and enforcement powers and composed of members drawn from the judiciary, the AI domain, and civil society. The SEBI model provides us with a regulatory template for a proposed independent supervisory authority. Like SEBI, such a body can be constituted through dedicated parliamentary enactment, with a defined statutory identity, enforcement capacity, and institutional accountability. This makes it entirely independent of the Supreme Court and the High Courts whose AI deployment it will supervise. Secondly, to bring transparency in the approval process for entry of private players in AI deployment, we can build a system based on the EU’s Conformity Assessment regime under Articles 43 – 49 of the AI Act, which mandates harmonized standards through third-party notified assessments in high-risk applications.  Thirdly, for the protection of litigants, the inspiration can be taken from the EU’s General Data Protection Regulation (GDPR), which grants data subjects the right to seek material and non-material compensation for data breaches.

 CONCLUSION

The Draft Regulations for Use of AI in Indian Courts 2026 constitute a noteworthy and significant step in the emerging global relationship between law and AI. But it can only be seen as a foundation rather than a final answer.  Yes, there are many flaws and loopholes. The success of these regulations will ultimately depend on the institutional will to implement them consistently across the vast judicial system of India. As AI systems grow more capable and less interpretable, the principle of judicial legitimacy must remain non-negotiable.

^[i] Sahaj Sankaran, AI in Courts: How India’s Draft Rules Stack Up Against the EU, US and China, ThePrint (June 5, 2026), https://theprint.in/judiciary/ai-in-courts-how-indias-draft-rules-stack-up-against-the-eu-us-and-china/2951484/.