ON JUNE 3, 2026, the Supreme Court of India’s Artificial Intelligence Committee released, about thirty-odd pages long, Draft Regulations for Use of Artificial Intelligence in Courts, 2026 (‘Draft Regulations’). These Draft Regulations mark the first in India’s history where the Supreme Court attempts to set rules governing the operation of artificial intelligence (AI) within Indian Courts.
The Draft Regulations were published for stakeholder and public comment, with a cut-off date of June 20, 2026.
What’s behind these Draft Regulations?
Until recently, there was no unified rulebook specifying which uses of AI in courts require approval, how oversight works, what permitted uses are, or what liability counts look like if errors occur.
The Draft Regulations aim to lay out a framework governing the use of AI in Indian courts. The five key principles that sit at the core of these draft regulations are: human primacy, transparency, accountability, data protection, and judicial independence. Each principle is evident in the provisions and embedded in the document’s architecture. The main point of the Draft Regulations appears to be that AI might help justice work better, but only when it stays under court control, is subservient, and never serves as a replacement.
The Draft Regulations will apply to the Supreme Court of India, all High Courts, and every court, tribunal, and statutory commission performing adjudicatory functions across the country. They will kick off when the dates are notified separately by the Supreme Court and each High Court.
The vocabulary of the draft regulations
Under Chapter I, the Draft Regulations build a definitional architecture and a careful vocabulary.
An “AI System” or “AI Tool” as defined under Regulation 3(1)(i) of the draft regulations is “any software, platform, application, device, or process” that employs AI to perform tasks in connection with any court process. The scope of this definition is broad by design, but it carefully excludes general-purpose software, such as word processors or speech-to-text processors, unless it specifically embeds or is functionally dependent on AI.
It separately defines “Adjudicatory Function” and “Administrative Function” (Regulation 3(1)(a) and 3(1)(b) respectively). That distinction is important and carries weight. Adjudicatory functions, such as hearing evidence, framing issues, pronouncing judgments and sentences, will attract a higher level of human oversight, whereas in terms of administrative functions, such as filing, scheduling, notice issuance, and record maintenance, they may allow for somewhat greater automation, though objectively always within sanctioned limits.
There are several definitions that address distinct risks and concepts. For example, Regulation 3(1)(z) defines “Hallucination” as the phenomenon by which an AI system produces outputs that appear plausible but are factually wrong, fabricated, or unsupported, including fabricated citations to cases or statutes. Regulation 3(1)(n) defines “Black Box” as an AI system whose internal reasoning is not transparent and cannot be explained in accessible terms. Regulation 3(1)(zg) defines “Risk Scoring” as the use of AI to assign a score estimating the probability of a person’s future behaviour, such as commission of an offence, recidivism, or failure to appear before a court. Regulation 3(1)(j) defines “Algorithmic Decision-Making” (‘ADM’) as any use of algorithmic output to inform, recommend, or arrive at a decision affecting a person or process.
“Human-in-the-Loop” (‘HITL’) is defined under Regulation 3(1)(zb). It is perhaps the most critical operational concept in the entire document. It means a governance process in which every AI output is subject to mandatory human review and where the final decision-making authority always vests with a human.
The definition of “explainability” as defined under Regulation 3(1)(w) notes that an AI system is explainable if it can, upon request, generate a comprehensible account of the reasoning and factors behind its output “in terms that can be understood by the judicial officer, court staff, or litigant” without requiring specialist technical knowledge from the recipient. This sets a demanding standard. Explainability is measured not by what the system can produce in principle, but by what a non-expert person can actually understand.
The Draft Regulations’ definitional architecture also defines both basic and advanced concepts, such as anonymisation, artificial intelligence, generative artificial intelligence, large language model, machine learning, sensitive judicial data, synthetic data/information, data minimisation, and more.
The governing principles
Chapter II sets out the general principles. The general principles, when read together, form a working guide for how India’s Courts can responsibly adopt, deploy and use AI.
What remains constant across is that humans will stay in charge, as noted under Regulation 4, “human primacy and judicial independence”. AI will not be allowed to take full control. Judges will have the full power to decide every case. AI can be used for assistance, whereas decisions about law, fact and justice belong entirely to the court. There are no exceptions here.
Regulation 5 notes the “Rule of Law” principle that AI adoption must comply with the Constitution and laws in force, including the principles of natural justice, without undermining due process, the right to a fair trial, equality before law, or access to justice. It states that judges should follow the 2002 Bangalore Principles of Judicial Conduct when addressing AI issues.
Regulation 6 notes that, to keep processes fair and non-discriminatory, an AI system should never reinforce unfair treatment based on traits such as “race, religion, caste, sex, gender, disability, language, economic status, or any other ground” prohibited under the Constitution or any law in force. While paying close attention to protecting “the rights and interests of vulnerable groups, including women, children, persons with disabilities, marginalised and minority communities, and persons from economically and socially disadvantaged backgrounds.”
Regulation 7 “Transparency and Explainability”- on use of AI systems in courts - high standards of transparency and explainability shall be met. Use of AI systems shall be explainable, while subject to higher scrutiny where “opaque or incapable”, and “shall be restricted in high-risk applications affecting personal liberty or any lawful right of a person.”
Regulation 8, “Accountability,” requires that every decision involving AI be treated as advisory and be attributed to the officer who made it. Further, in no case can accountability be avoided, such as for reasons of a black box or hallucination.
Regulation 9 mandates “continuous monitoring and periodic technical, legal and ethical audits of AI systems” and requires that they be recorded and disclosed when required by the regulations. Regulation 10 requires data to be protected in accordance with the provisions of the Digital Personal Data Protection Act, 2023, or any other law in force. Notably, sensitive judicial data, as defined under Regulation 3(a)(zh), shall be protected to the highest standard.
Regulation 11 “Purpose limitation” mentions that each AI system’s use must stay locked to its original purpose only, and any shifting in purpose will need fresh approval, with reasons in record. Regulation 12, “Proportionality,” addresses how much protection is appropriate for the risk, which determines how strong the guardrails must be. When personal liberty or significant legal rights are at stake, HITL is mandatory with independent oversight. Regulation 13 “Inclusivity and accessibility” requires that AI systems promote inclusivity and be equitable, with attention to fair access among stakeholders, including those “from rural, economically disadvantaged, or linguistically diverse communities,” and help address the digital divide.
Regulations 14 and 15 talk about the legality and accuracy of data and appropriate cybersecurity measures, respectively.
The last two principles deserve particular attention. Regulation 16 “Presumption in favour of responsible AI adoption” requires courts to consider opportunities to use AI when it helps “improve access to justice, reduce delays, or enhance administrative efficiency”, without losing the human decision-making primacy and requires recording reasons in writing when an AI system is restricted or refused for use. The second one, under Regulation 17 “Innovation over restraint”, requires that AI be adopted in courts responsibly and actively, rather than inaction.
Permissible uses of AI systems in courts
Chapter III (Regulations 18-21) sets boundaries on the permissible and prohibited uses of AI systems in courts. One cannot relax the prohibited list; it is non-derogable.
The draft regulations on permissible uses of AI systems (Regulation 20), against prior written approval and designated human supervision, include: case management (including identification of defects in new filings), cause list preparation, hearing scheduling and docket prioritisation; automated transcription of proceedings, with mandatory review and accuracy certification; translation of judgments, orders, pleadings and other legal document, subject to human verification; legal research, precedent retrieval, citation verification, and document summarisation; administrative functions including case filing assistance, defect scrutiny, record management and judicial resource allocation; conversational AI chatbots to assist stakeholders in accessing court services; accessibility tools for persons with disabilities or language barriers; document authenticity verification and fraud detection in administrative processes, with mandatory human review of all outputs; anonymisation of court records for public domain publication; analytical tools for court performance assessment and backlog management; and auto-generation of prescribed notices and summons.
The above list is “illustrative and not exhaustive”. Any use not on this list requires specific prior authorisation from the Appropriate Authority.
The prohibited uses of AI systems (Regulation 20) are “absolute and non-derogable” and cannot be relaxed or modified, including: Personal data cannot be taken in use to train or test any AI system without prior approval, and compliance with applicable data protection law; No judicial outcome shall be reached through ADM alone; No AI system shall perform adjudication or sentencing without mandatory HITL; Risk Scoring is entirely prohibited; No undisclosed or unexplainable AI system shall be used in any proceeding that could materially affect a party’s legal rights or personal liberty; No AI system may profile or infer the future conduct of parties, accused persons, witnesses, or lawyers; No AI may be used for the surveillance or continuous monitoring of judicial officers, advocates, or litigants, except as specifically authorised by law; No AI-generated output may be submitted as independent evidence without full disclosure; and no AI system may be used in any manner that confidentiality of judicial deliberations or the independence of the judicial decision-making process.
The institutional architecture
Chapter IV (Regulation 22-34) of the draft regulations, titled “Policy Making and Institutional Mechanism”, envisions a layered governance structure, stretching down from the Supreme Court to every High Court, tribunal, and district court.
Regulation 22 envisages “a permanent, full-time Apex Body” with its seat at the Supreme Court. It will be chaired by a Supreme Court judge nominated by the Chief Justice of India, and its membership includes two Chief Justices of High Courts, two High Court judges, a representative from an institution of national importance, a Joint Secretary from the Ministry of Electronics and Information Technology, and experts in finance, cybersecurity, technology law, and data privacy. They will also be joined by a professor heading AI at the National Judicial Academy as a member.
The functions of the Apex Body (Regulation 23) include ensuring that no AI systems violate any provisions of the Constitution or any law in force; To standardise tests and unify innovative AI systems across the courts in India; Upholding governing principles (Chapter II); following the vision of the AI Committee; Coordinating with High Court AI Committees for a harmonised approach to AI governance across all courts; To liaise with statutory or regulatory body on matters concerning AI in courts; Guiding actions in respect of grievances against AI systems; and Publish an Annual Governance Report.
Regulation 24 guides the power of the Apex Body; it sets minimum mandatory standards for AI use across all courts in India; Approves AI systems; Issues guidelines for implementation of these draft regulations; Oversees the Centre of Research and Excellence on AI; and Constitutes Standing Committees to help discharge its functions.
The Apex Body is followed by five committees to assist in the discharge of its functions (Regulation 25): a Judicial Committee; a Technical Committee; a Committee on Infrastructure and Finance; a Case and Data Management Committee; and a Cyber Security Committee (as outlined in Regulations 26-31, respectively).
An integrated body, Centre of Research and Excellence on AI (CoRE-AI), will be established under Regulation 32, comprising Judges, lawyers, technologists, and academics. It is conceived as the judiciary’s intellectual infrastructure for understanding AI. It will conduct original research, evaluate AI tools, maintain a centralised record of AI tools, evaluations and technical regulations, advise and support the Apex Body and its committees, publish white papers and empirical studies, and organise conferences/similar events, etc.
Under Regulation 33, the Supreme Court and every High Court are to establish their own AI Committee to govern the adoption of AI within their jurisdictions. Regulation 34 requires the support of every AI Committee by a dedicated AI Secretariat.
Oversight, accountability, and what happens when things go wrong
Regulations 35-45 under Chapter V address oversight, audits, and incident management.
A comprehensive Technical and Ethical Impact Assessment must be submitted to the Appropriate Authority before any AI System may be deployed in a court process. The assessment shall cover the purpose, architecture, and functioning of the AI system; the nature, source, quality and representativeness of its training data; risks of bias, error, hallucination and misuse in the relevant judicial context; cyber security vulnerabilities and the data protection measures in place; mechanisms for explainability and compliance with Human-in-the-Loop requirements; and procedures for redressal of any harm and for incident reporting.
In suitable cases, the Appropriate Authority may direct any AI system deployed in a court process to be evaluated through Controlled Environment Testing. The testing shall be undertaken on a clearly defined and time-limited basis, and the parameters of evaluation and outcomes to be recorded for consideration presented before the Appropriate Authority in relation to the deployment of the AI system.
An AI register must be maintained in each court, consisting of particulars, including: a record of all AI systems approved for use; the purposes and conditions of approval; the identity of the AI Service Provider; records of impact assessments and audits; and all AI incidents associated with each system. Subject to data protection, confidentiality, and cybersecurity considerations, the register shall be uploaded to the court's official website for public access.
An in-house periodic technical, legal, and ethical audit shall be conducted on all court AI systems and submitted to the Appropriate Authority, with the results recorded in the AI register. An AI incident database will be maintained at every AI secretariat, consisting of systematic recording of all AI incidents, including their type, cause, manner of occurrence, consequences and the remedial measures taken. Further, where an AI incident occurs in any High Court, the takeaways from it must be communicated to the AI secretariats of all other High Courts and to the Apex Body.
In consultation with the AI secretariat, every High Court shall establish and maintain an emergency and fallback protocol specifying the procedure to be followed in the event of a failure, malfunction, or unavailability of any AI system in court.
On the note of transparency and disclosure, the court shall inform parties when AI materially assists in case management, document analysis, or judicial administration that may affect the conduct of their proceedings. Litigants and their legal representatives, when used AI in preparation or summation of any document, must disclose to the court at the time of submission. Additionally, the Court shall have the authority to require disclosure of the AI System used, the nature and extent of AI assistance provided, and the steps taken to verify the accuracy of any AI-generated content, in respect of any AI-assisted submission placed before it.
Synthetic data/information usage by any person shall be disclosed to the court. In the instances where the AI-generated character in any document, pleading, or evidence submitted to the court is found to be fabricated, false, misleading or inaccurate by reason, the responsibility shall be borne by the person submitting the same, and that person shall not be entitled to rely upon the character of the AI output as a defence.
An AI content verification authority shall be constituted to oversee the operation and continuous updation of verification standards, tools and protocols applicable to GenAI-generated content in court process. An AI adoption annual Transparency Report shall be submitted by every High Court, Tribunal and Commission falling under these draft regulations, summarising the AI systems in use, outcomes of audits, AI incidents recorded and measures taken for compliance with and improvement of these draft regulations, to the apex body and publish the same on its official website.
Private sector engagement, data protection, and capacity building
Chapter VI (Procurement and Private Sector Engagement) – Regulation 46 - The engagement of Private Entities is tightly regulated. Subject to “a comprehensive evaluation covering technical capability, legal compliance, ethical standards, data security practices and financial standing”, they “shall undertake, participate in, or provide any service in connection with an AI System deployed” in court processes only with written approval from the Appropriate Authority. AI-related services agreements entered with private entities must satisfy the mandatory provisions as required under Regulation 46(4)(a-l).
Ownership stays with courts when AI systems grow from legal data or tools paid for by the judiciary. Even if outside groups build something new, they cannot lock up results behind proprietary claims. A lasting, cost-free access right remains with the court system. What emerges from public case records belongs to the people's justice framework first. Private firms might help shape these tools, but never fully own them. Where AI tools are developed using court data or court resources, “the court retains ownership of, or a perpetual royalty-free license to the resulting tool and its outputs.”
Chapter VII – (Data Protection and Cyber Security) – Regulations 47-48 - the draft regulations require compliance with the Digital Personal Data Protection Act, 2023, the Information Technology Act, 2000 and the applicable rules and regulations, and any other law governing the protection of personal data and judicial information for the time being in force.
“Sensitive judicial data shall not be transferred to any external system without express written authorisation”. All transfers “shall be subject to appropriate technical and contractual safeguards designed to prevent unauthorised access, disclosure, alteration, or misuse.” Systems processing sensitive judicial data must use on-premise or sovereign cloud deployment. Data minimisation, anonymisation before training, and strict access controls on a need-to-know basis are mandatory. “Every AI System in use in court processes shall be subject to regular cybersecurity audits.”
Chapter VIII – (Capacity Building, Training, and Best Practices) – Regulations 49-51 - requires that all judges, advocates, and court staff who interact with AI systems receive regular, structured training on the technical, legal, and ethical dimensions of AI. This training must be accessible to personnel in district courts, offered in a manner that accounts for linguistic diversity. The training programmes shall address followings as outlined under Regulation 49(3)(a-e) and shall be periodically reviewed.
Grievance redressal and the right to seek remedy
Chapter IX – (Grievance Redressal and Remedies) – Regulations 52-53 – Where a prohibited use of AI in a court proceeding harms any party, that party or their legal representative may file an application before the respective court where the AI was used. The court, after hearing both sides, passes such orders as it thinks fit. High Courts may lay down principles and procedures for adjudicating such grievances and may prescribe formats to ensure accessibility for persons with limited legal literacy. Further, nothing in the regulations forecloses any other legal remedy available under any existing law.
What this framework is doing, and what it is not
The Draft Regulations are not anti-technology. The ‘Presumption in Favour of Responsible Adoption and the Innovation over Restraint’ principle makes plain that the Supreme Court’s AI Committee views AI as something to be embraced with care, not refused on principle. The document creates an enabling structure, not a prohibition regime.
Every regulation that expands AI use in courts is paired with an anchoring accountability to a human, a transparency obligation to the public, or an absolute prohibition that no authority may relax. The ban on Risk Scoring is perhaps the clearest statement of what the Supreme Court is refusing.
The Apex Body, the AI Committee, the AI Secretariat, and the CoRE-AI will need funding, staffing, and a culture that takes compliance seriously rather than treating it as a formality. The requirement that audits be conducted in-house, and that source code never be shared with external parties, assumes a level of internal technical capacity that will take time to build.
There are also questions the draft regulations leave open. For instance, how will the regulations interface with the emerging AI governance landscape at the national level? What happens when AI-generated content is submitted by a party that did not know it was hallucinated? How will the grievance mechanism function in practice for a litigant in a district court with no legal representation?
What we can only expect of the Draft Regulations, as of now, is that it makes expectations clear, violations visible, and accountability inescapable.
To read our analysis of the Supreme Court’s Draft AI Regulations, click here.
