Inside SC’s proposed regulations for AI use in courts: What’s allowed, what’s absolutely barred
An AI Committee constituted by the Supreme Court released the draft Regulations for Use of Artificial Intelligence in Courts, 2026, on 3 June 2026, for public consultation.
The draft regulations set out permitted uses of AI in courts (legal research, drafting assistance, transcription, translation, scheduling, case management, record maintenance) alongside a set of absolute prohibitions on AI performing adjudicatory functions.
The draft proposes creation of a full-time apex body at the Supreme Court to set standards, approve AI systems for judicial use, and oversee compliance across all courts, including subordinate courts and statutory tribunals.
Public comments were initially invited until 20 June 2026; the Court has since extended the consultation window to 15 July 2026, ahead of finalisation.
Judicial Independence and the Limits of AI in Adjudication
Judicial independence is a basic feature of the Constitution (as held by the Supreme Court), requiring that judicial decision-making remain free from external or mechanical determination. The draft regulations operationalise this principle for the AI era by treating any AI output used in court as strictly advisory, never as a substitute for the mind of the judge. This tracks the constitutional expectation, drawn from the separation of powers and Article 50 (separation of judiciary from executive) read with the basic structure doctrine, that adjudication is an exercise of reasoned human judgment, not automated output.
Key Details
- The draft absolutely bars AI from reaching a judicial outcome alone, deciding bail, determining sentences, or evaluating witness credibility.
- "Human-in-the-loop" is made mandatory wherever AI assists any process that could affect a party's legal rights or personal liberty.
- AI-generated output is treated as advisory only; courts and parties must independently verify its accuracy before relying on it.
The specific bars on bail decisions, sentencing, and witness-credibility evaluation are the clearest application of the "advisory-only" principle — they mark the boundary between permissible administrative/research assistance and impermissible delegation of judicial function to a machine.
Natural Justice, Due Process, and Algorithmic Accountability
Principles of natural justice — the right to a fair hearing and reasoned decision, part of the due process guarantee under Article 21 as expanded by Maneka Gandhi v. Union of India (1978) — require that decisions affecting rights be explainable and open to challenge. The draft regulations extend this to algorithmic tools: any undisclosed or unexplainable AI system cannot be used where it could materially affect a party's legal rights or personal liberty, and parties relying on AI-assisted material must disclose its use and the verification steps taken.
Key Details
- Mandatory disclosure: any party using AI to prepare pleadings, submissions, or evidence must disclose this at the time of filing.
- "Risk scoring" tools (e.g., predicting recidivism or flight risk) are entirely prohibited.
- AI-generated material cannot be submitted as independent evidence without full disclosure of its provenance and verification.
- Profiling or predicting the future conduct of parties, accused persons, witnesses, or lawyers by AI is barred, as is AI-based surveillance of judicial officers, advocates, or litigants (absent specific legal authorisation).
The transparency and disclosure requirements mirror the due-process logic of Maneka Gandhi — a decision (or evidence) that cannot be explained or scrutinised cannot fairly bind a litigant, so any AI role in a proceeding must be visible and contestable.
Institutional Oversight: The Proposed Apex AI Body
The draft proposes a full-time apex body located at the Supreme Court to frame standards and guidelines for AI adoption across the judiciary, vet and approve specific AI systems before courts can use them, and monitor compliance. This is consistent with the Supreme Court's e-Committee model, which has historically driven ICT-based judicial reforms (e-Courts Mission Mode Project) — the AI regulations extend that institutional oversight approach from digitisation to AI governance specifically.
Key Details
- The draft regulations were prepared by an AI Committee of the Supreme Court headed by a senior judge, with several other judges as members.
- Scope extends beyond the Supreme Court to all courts — subordinate courts, district-level tribunals, and statutory adjudicatory bodies.
- The proposed apex body's approval function means individual courts cannot independently adopt any AI tool without central vetting, a centralised-standards approach.
Establishing a dedicated apex body reflects the broader Indian regulatory pattern of creating specialised institutional mechanisms (comparable to sectoral regulators) to govern a new technology domain, ensuring uniform standards rather than fragmented adoption of AI tools by individual courts.
- Draft Regulations for Use of AI in Courts, 2026 released for public consultation on 3 June 2026.
- Original comment deadline: 20 June 2026; extended to 15 July 2026.
- Permitted AI uses: legal research, citation verification, drafting assistance, translation, transcription, scheduling, case management, record maintenance, judicial administration.
- Absolutely barred: AI-only judicial outcomes, AI-based bail/sentencing decisions, risk-scoring/profiling of parties or witnesses, undisclosed/unexplainable AI systems affecting rights or liberty, AI-based surveillance of judicial officers/advocates/litigants without legal authorisation.
- Mandatory human-in-the-loop for any AI-assisted process capable of affecting a party's legal rights or personal liberty.
- Draft proposes a full-time apex body at the Supreme Court to set standards and approve AI systems used across all courts, including subordinate courts and statutory tribunals.