The Hon’ble Supreme Court of India, under the aegis of its Artificial Intelligence Committee, has published the “Draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026” and has invited comments and suggestions from stakeholders and the general public. The proposed regulations seek to establish a comprehensive framework governing the adoption, deployment, integration, oversight and governance of AI systems across courts, tribunals and statutory commissions performing adjudicatory functions in India. The framework is founded on the principles of human primacy, judicial independence, transparency, accountability, data protection, cybersecurity and responsible AI adoption.

 

Definitions (Regulation 3)

Regulation 3 introduces and defines several key technical and governance-related terms that form the foundation of the regulatory framework. The Regulation seeks to establish a common understanding of concepts associated with Artificial Intelligence and its deployment within court processes.

Among the important definitions, “Algorithmic Decision-Making (ADM)” refers to the use of outputs produced by algorithms to inform, recommend, or arrive at a decision affecting any person or process. A “Black Box AI System” is an AI system which employs deep learning techniques and whose internal logic, decision-making process cannot be explained by reference to identifiable rules or factors accessible to the user or affected person. “Hallucination” means the generation of information, outputs, or content by an AI system that appears plausible but is factually incorrect, fabricated, misleading, or unsupported by verifiable data or source material. “Human-in-the-Loop (HITL)” refers to a governance and operational model requiring meaningful human oversight, review, intervention, or control over AI system outputs before they are acted upon or relied upon. The Regulation also defines “Appropriate Authority” as the Apex Body at the Hon’ble Supreme Court or the AI Committee at the respective Hon’ble High Court or Tribunal or Commission. Further, “Sensitive Judicial Data” means personal identifiable information of parties, witnesses, or legal representatives in connection with a court process, whose unauthorized disclosure may cause harm.

 

Guiding Principles Governing AI Adoption (Regulations 4-17)

Regulations 4 to 17 lay down the foundational principles governing the adoption, deployment and use of AI Systems in courts. Regulation 4 mandates that AI shall remain strictly assistive in nature and function solely in an assistive capacity without supplanting or compromising the independent exercise of judicial authority. The ultimate authority to determine matters of law, fact and justice continues to vest exclusively in judicial officers.

Regulations 5 to 8 require adherence to the Constitution, principles of natural justice, fairness, non-discrimination, transparency, explainability and accountability. AI-generated outputs are to be treated as advisory in nature and accountability for decisions made with the assistance of AI remains exclusively with the concerned officer.

Regulations 9 to 15 provide for continuous monitoring; periodic technical, legal and ethical audits; protection of personal data and sensitive judicial data, purpose limitation, proportionality, inclusivity, accessibility, data integrity and cybersecurity safeguards.

Significantly, Regulations 16 and 17 create a presumption in favour of responsible AI adoption and encourage innovation in the justice delivery system. However, they expressly provide that AI systems shall not replace humans in decision-making and shall not be deployed for dispute-outcome prediction.

 

Permissible Uses of AI (Regulation 19)

Regulation 19 permits the use of AI Systems, subject to prior written approval of the Appropriate Authority and supervision/verification by nominated officers, for various administrative and assistive functions. These include case management, identification of defects in filings, cause list preparation, hearing scheduling and docket prioritisation, automated transcription of court proceedings, translation of judgments and legal documents, legal research, precedent retrieval, citation verification and document summarisation.

The regulation further permits AI-assisted case filing, defect scrutiny, record management, judicial resource allocation, litigant assistance through conversational AI assistants and guided chatbots, accessibility services for persons with disabilities and language barriers, document authenticity verification, fraud detection, anonymisation of judgments and court records, judicial administration, court performance assessment, backlog management and auto-generation of prescribed formats, notices and summons.

 

Prohibited Uses of AI & Violations (Regulations 20-21)

Regulation 20 prescribes absolute and non-derogable prohibitions on specified uses of AI in court processes. Personal data cannot be used to train, test or refine AI systems without prior approval of the Appropriate Authority and compliance with applicable data protection laws. AI cannot independently determine judicial outcomes, perform adjudication or sentencing functions, or be used through Algorithmic decision-making alone in adjudicatory matters. Any AI output in such matters must remain advisory and be subject to independent judicial evaluation.

The regulation further prohibits Risk Scoring, including assessment of flight risk, prediction of recidivism (person’s relapse into criminal behavior), evaluation of bail eligibility and determination of credibility of parties or witnesses. It also prohibits: i) prediction of future conduct, ii) use of opaque or unexplainable AI systems affecting lawful rights or personal liberty, iii) surveillance of judges, advocates or litigants, iv) submission of AI-generated material as evidence without disclosure of its AI-generated character, and v) any use that may compromise judicial deliberations or judicial independence.

Under Regulation 21, every violation of Regulation 20 must be reported to the AI Secretariat and placed before the AI Committee, which may direct appropriate remedial measures, including suspension of the relevant AI System.

 

Institutional Framework and Governance Structure (Regulations 22-34)

Regulations 22 to 34 establish a comprehensive institutional mechanism for AI governance within the judiciary. Regulation 22 creates a permanent Apex Body at the Supreme Court level comprising judges, technical experts, cybersecurity specialists, finance experts, representatives from Ministry of Electronics and Information Technology, members of academia and members of the Bar.

Regulations 23 and 24 define the functions and powers of the Apex Body, including formulation of minimum mandatory standards, approval of AI systems, policy development, coordination with High Courts and oversight of AI governance at the national level.

Regulations 25 to 32 establish specialised committees, namely the Judicial Committee, Technical Committee, Committee on Infrastructure and Finance, Case and Data Management Committee and Cyber Security Committee, and the Centre of Research and Excellence on Artificial Intelligence (CoRE-AI), which will undertake research, evaluation, monitoring of technological developments and policy support.

Regulations 33 and 34 require constitution of AI Committees and AI Secretariats in the Supreme Court and every High Court to oversee implementation, approvals, audits, compliance monitoring and administration of AI systems within their respective jurisdictions.

 

Oversight, Audits and Transparency Measures (Regulations 35-45)

Regulations 35 to 45 establish an extensive oversight and accountability framework. Regulation 35 mandates submission of a comprehensive Technical and Ethical Impact Assessment before approval of any AI System for use in Court processes. Such assessment must evaluate, inter alia, the system architecture, training data, risks of bias, error and hallucination, cybersecurity vulnerabilities, explainability mechanisms and Human-in-the-Loop compliance (all such specific terms defined under the said Regulations).

Regulation 36 provides for Controlled Environment Testing before full-scale deployment of AI systems. Regulations 37 to 39 require maintenance of AI Registers, periodic in-house technical, legal and ethical audits and creation of AI Incident Databases for recording and addressing AI-related incidents.

Regulations 40 to 42 preserve human discretion over AI-generated outputs, require independent professional judgment by supervising officers and mandate emergency and fall-back protocols to ensure continuity of court processes in case of AI failure or suspension.

Regulation 43 requires transparency and disclosure whenever AI tools materially assist court processes. Parties and advocates using AI in preparing documents, pleadings or evidence must disclose the AI-assisted character of such material by filing a declaration or certificate in the prescribed format. The Court may require disclosure of the AI system used, the nature of assistance provided, and the steps taken to verify accuracy. Where AI-generated content is found to be fabricated, false, misleading or inaccurate, the person submitting it shall bear full responsibility and may be subjected to appropriate orders by the Court.

Regulations 44 and 45 provide for constitution of an AI Content Verification Authority to oversee verification standards for GenAI-generated content and require publication of Annual Transparency Reports detailing AI adoption, audit outcomes, AI incidents and compliance measures.

 

Private Sector Engagement and Data Protection (Regulations 46-48)

Regulation 46 governs engagement of private entities, vendors and AI service providers. Such engagement requires prior approval of the Appropriate Authority and is subject to evaluation of technical capability, legal compliance, ethical standards, cybersecurity safeguards and financial standing. Agreements with private entities must contain provisions relating to ownership of court data, restrictions on use of judicial data, audit rights, disclosure obligations, source and model transparency, explainability documentation, indemnity obligations and allocation of liability. The said regulation prohibits retraining, fine-tuning or modification of AI models using court data without express written approval. Further, the courts shall retain ownership of, or a perpetual royalty-free licence to, AI tools developed using court data or court resources.

Regulations 47 and 48 mandate compliance with the Digital Personal Data Protection Act, 2023, the Information Technology Act, 2000 and other applicable laws. Sensitive judicial data cannot be transferred to external systems without express written authorisation and must be protected through data minimisation, anonymisation, access controls and periodic cybersecurity audits.

 

Capacity Building and Grievance Redressal (Regulations 49-53)

Regulations 49 to 51 require regular and structured training programmes for judges, advocates and court staff on the technical, legal and ethical dimensions of AI. Training must address AI capabilities and limitations, bias, hallucinations, technical errors, cybersecurity, data protection obligations, and reporting of AI incidents. The regulations also provide for maintenance of a repository of best practices and periodic review of training programmes.

Regulations 52 and 53 establish grievance redressal mechanisms. Where any harm is caused as a direct or indirect consequence of a prohibited use of AI under Regulation 20, the affected person or his legal representative may approach the concerned Court, which may pass appropriate orders after providing an opportunity of hearing. The regulations also preserve all other remedies available under applicable law.

 

CONCLUSION

The Draft Regulations constitute the first comprehensive framework for AI governance within the Indian judicial system. While actively encouraging responsible AI adoption and technological innovation in court administration, the framework firmly preserves human control over adjudicatory functions and establishes safeguards relating to transparency, accountability, privacy, cybersecurity, judicial independence and protection of litigants’ rights.