On 3 June 2026, the AI Committee of the Supreme Court of India published a preliminary draft of the Regulations for Use of Artificial Intelligence in Courts, 2026 (the “Draft Regulations”), inviting comments from all stakeholders and the general public by 20 June 2026. The draft represents the most significant institutional step India has taken toward formalising the governance of AI within its judicial system.

Background and Context

The timing of the Draft Regulations reflects a broader international trend. Courts across multiple jurisdictions have begun grappling with the risks and opportunities that generative AI tools introduce to legal proceedings, from the filing of AI-drafted pleadings containing fabricated citations to the potential for algorithmic systems to influence sentencing and bail determinations. The Draft Regulations attempt to address this landscape comprehensively, drawing on principles of judicial independence, data protection, transparency, and proportionality to construct a governance architecture that spans the Supreme Court down to district-level courts and statutory commissions.

The regulatory effort builds on the existing work of the National Court Management System and aligns with India’s existing legislative framework, including the Digital Personal Data Protection Act, 2023 and the Information Technology Act, 2000, without displacing either.

Structure and Architecture

The Draft Regulations are organised across ten chapters. Chapter I provides definitions and determines scope. Chapters II and III establish governing principles and draw the line between permissible and prohibited uses. Chapters IV and V address institutional governance and oversight. The remaining chapters cover procurement from private entities, data protection, capacity building, grievance redressal, and general provisions.

Several definitional choices merit attention. “AI System” and “AI Tool” are defined broadly to include any software, platform, or process that employs artificial intelligence in connection with any court process, but are expressly carved out to exclude general-purpose digital tools unless specifically embedded with, or functionally dependent upon, AI. The definitions of “hallucination,” “Large Language Model,” “Generative Artificial Intelligence,” and “Black Box” are each defined with technical precision unusual for a court document, signalling a degree of AI literacy in the drafting process that is itself notable.

Governing Principles: The Normative Core

Chapter II enumerates fourteen governing principles that collectively frame the philosophy of the Draft Regulations. Human primacy is foregrounded: every AI system must function solely in an assistive capacity, and no AI system may supplant or compromise independent judicial authority. This is not merely declaratory. The principle is operationalised throughout the document through mandatory Human-in-the-Loop requirements for any application involving personal liberty, the absolute prohibition on algorithmic adjudication, and accountability rules that vest responsibility personally in the relevant officer regardless of what any AI system may have generated.

The principle of fairness and non-discrimination expressly prohibits the deployment of any AI system that perpetuates, amplifies, or introduces bias on grounds of race, religion, caste, sex, gender, disability, language, or economic status. This aligns the draft with Articles 14 and 15 of the Constitution of India, and reflects growing concern globally that training datasets frequently embed historical social biases that, if unchecked, could systematically disadvantage already marginalised litigants.

Two principles that sit in creative tension deserve note. Regulation 16 establishes a “Presumption in Favour of Responsible AI Adoption,” stating that courts should actively seek opportunities to deploy AI that demonstrably improves access to justice, and that refusals to permit AI use must be justified in writing. Regulation 17 reinforces this as “Innovation over Restraint,” expressly encouraging active adoption over caution where other principles are otherwise satisfied. These provisions appear designed to counteract institutional inertia, though they operate subject to the absolute prohibitions in Chapter III, which cannot be overridden by any authority.

Permissible and Prohibited Uses

Chapter III is the most immediately consequential chapter for practitioners and technology vendors. Regulation 19 sets out a list of permissible uses that is expressly illustrative and non-exhaustive. These include case management and cause-list preparation; automated transcription (with mandatory human review); translation of judgments and pleadings; legal research, precedent retrieval and citation verification; AI-powered chatbots to assist litigants; accessibility tools for persons with disabilities; document authenticity verification and fraud detection; anonymisation of judgments for publication; and auto-generation of prescribed notices and summons.

Regulation 20 sets out absolute prohibitions that are described as non-derogable and not subject to relaxation by any authority, including the Chief Justice. The key prohibitions are:

• Reaching any judicial outcome through Algorithmic Decision-Making alone, or solely on the basis of AI-generated information or analysis
• Any AI system performing adjudication or sentencing without mandatory Human-in-the-Loop supervision
• Risk Scoring for any purpose, including assessment of flight risk, prediction of recidivism, evaluation of bail eligibility, or assessment of the credibility of parties or witnesses
• Profiling or predicting the future conduct or behaviour of parties, accused persons, witnesses, or legal representatives
• Surveillance or continuous monitoring of judicial officers, advocates, or litigants within court premises or in connection with court processes
• Submitting AI-generated output as an independent source of evidence without full and transparent disclosure of its AI-generated character
• Use of personal data to train, test, or refine any AI system without prior approval and compliance with data protection law
• Any use that compromises the confidentiality of judicial deliberations or the independence of judicial decision-making

The prohibition on risk scoring is particularly noteworthy. It represents an unambiguous rejection of the use of predictive algorithms in bail and sentencing decisions, at a time when such tools have been controversially adopted in several other jurisdictions. The decision to make this prohibition absolute and non-derogable places India in a distinct position in the comparative regulatory landscape.

Institutional Framework

Chapter IV establishes a multi-tier governance structure. At the apex sits a permanent, full-time Apex Body at the Supreme Court, chaired by a Judge of the Supreme Court nominated by the Chief Justice of India, with membership drawn from High Court Chief Justices and Judges, an officer from the Ministry of Electronics and Information Technology, and experts in finance, cybersecurity and technology law. The Apex Body is charged with standard-setting, benchmarking, coordination across High Courts, liaison with CERT-In and the Data Protection Board, and the publication of an Annual Governance Report.

Below the Apex Body, each High Court is required to constitute its own AI Committee, which serves as the Appropriate Authority within its jurisdiction. AI Committees are supported by a dedicated AI Secretariat, headed by an officer in the cadre of District Judge. The draft also provides for the establishment of a Centre of Research and Excellence on Artificial Intelligence (CoRE-AI), tasked with original research, evaluation of AI tools, comparative legal tracking, and the publication of white papers and empirical studies.

Oversight, Audits, and Incident Management

Chapter V introduces a detailed oversight regime. Every AI system proposed for court use must first undergo a Technical and Ethical Impact Assessment evaluating, at minimum, the system’s purpose and architecture, the nature and quality of its training data, risks of bias and hallucination, cyber security vulnerabilities, and explainability mechanisms. The Appropriate Authority is required to prescribe a standard format for this assessment within six months of commencement.

All court AI systems must undergo periodic audits at intervals not exceeding one year. A critical provision restricts audits to in-house conduct: source code, algorithms, datasets, and architectural information may not be shared with any third party or private entity for an audit conducted outside court premises. This constraint may prove practically challenging and is likely to attract comment during the consultation period.

Regulation 43 imposes mandatory disclosure obligations. Where any party or legal representative uses AI in the preparation of pleadings, evidence, or submissions, a declaration in a prescribed format must be filed at the time of submission. Submissions found to be fabricated, false, or misleading by reason of their AI-generated character attract full personal liability and may not be defended on the basis of the AI system’s output.

Procurement and Data Protection

Chapter VI imposes stringent conditions on private sector engagement. No vendor may participate in any AI system deployed in court processes without prior written approval. Contracts must address ownership of court data and AI outputs, prohibition on reuse of sensitive judicial data beyond the scope of engagement, indemnity clauses protecting the court from liability for vendor-supplied system defects, source code transparency requirements, and explicit prohibition on retraining AI models using court data without express committee approval. Where tools are developed using court data or resources, the court must retain ownership or a perpetual royalty-free licence.

Chapter VII requires compliance with the Digital Personal Data Protection Act, 2023 and the Information Technology Act, 2000. Sensitive judicial data may not be transferred to any external system without express written authorisation. The data minimisation principle is mandated at the design, procurement, deployment, and audit stages, and anonymisation must be applied before any personal data is used to train or test an AI system.

Implications for Legal Practitioners

Practitioners will need to pay close attention to the disclosure regime introduced by Regulation 43. The obligation to file a prescribed declaration whenever AI has materially assisted in the preparation of any document, pleading, or evidence applies at the time of submission. Law firms and in-house legal teams will need to establish clear protocols for tracking and documenting AI usage in matter preparation to ensure the disclosure threshold is consistently identified.

The accountability regime in Regulation 8 is equally significant. A judicial or administrative officer who uses AI assistance remains personally accountable for the correctness of any decision and may not invoke the opaqueness of a Black Box system, or the occurrence of hallucination, as a defence for an incorrect or harmful outcome. This principle reinforces the importance of independent verification of AI-generated outputs before reliance, and applies by extension to legal practitioners through the liability provision in Regulation 43(6).

Implications for Technology Vendors

For AI service providers, the Draft Regulations represent a materially heightened compliance threshold. Prior written approval is required before any engagement with court AI systems. Technical and ethical impact assessments, audit cooperation, incident reporting, and contractual restrictions on data use and retraining are baseline requirements. The prohibition on sharing source code with third-party auditors may require vendors to support in-house audit processes directly, with consequent implications for how contracts and service level agreements are structured.

Concluding Observations

The Draft Regulations represent a careful and detailed attempt to create a principled governance framework for AI in India’s judicial system. The document is notable for the clarity with which it draws the line between permissible and prohibited uses, and for the intellectual care with which key concepts are defined. The absolute prohibition on risk scoring and algorithmic adjudication reflects a clear normative position consistent with a rights-based approach to judicial governance, and distinguishes India’s proposed framework from jurisdictions that have been more permissive in this space.

Several provisions are likely to attract comment during the consultation period. The restriction of audits to in-house processes raises practical questions about technical capacity and independence. The requirement for AI Committee approval before any new AI use not enumerated in Regulation 19 may create administrative bottlenecks if not supported by streamlined approval processes.

Stakeholders have until 20 June 2026 to submit written comments to the Member Secretary, AI Committee, Supreme Court of India at office.regcc@sci.nic.in. Given the breadth and technical complexity of the draft, legal practitioners, technology companies, civil society organisations, and academic institutions with an interest in AI governance would be well advised to engage with the consultation process. The Draft Regulations, when finalised, are likely to set a significant benchmark not only for India’s judicial system but for AI governance frameworks in courts across the common law world.