Description

The Indian Society of Artificial Intelligence and Law (ISAIL) is pleased to submit its feedback on the draft Regulations for Use of Artificial Intelligence in Courts, 2026 developed and disseminated by the honourable Supreme Court of India.

Stakeholders

Hamda Arfeen, Co-Chairperson for the Policy Innovation Committee, Indian Society of Artificial Intelligence and Law

Nisha Singh, Co-Chairperson for the Policy Innovation Committee, Indian Society of Artificial Intelligence and Law

Areena Kausar, Regional Representative Member, Lucknow (Central India), Indian Society of Artificial Intelligence and Law

Prathik Karthikeyan, Regional Representative Member, Delhi NCR (Northern India), Indian Society of Artificial Intelligence and Law; Chief of Legal Engineering & Growth, LITT

Arshita Anand, Co-Founder, Vaquill.AI

Priyansh Khodiyar, Founder, Vaquill.AI

Rupal Mathur, Co-Founder, Rhett.legal

Swastik Grover, Co-Founder, Rhett.legal 

Detailed Feedback

These comments respond to the Notice dated 3 June 2026 on feedback for “Regulations for Use of Artificial Intelligence in Courts, 2026”. They are offered in support of the initiative and in the hope of strengthening it.

The draft is a serious and welcome instrument. It is rightly confined to the judiciary’s own use of artificial intelligence in court processes, and not to the wider market for legal technology or to the work of the Bar. On the question that matters most, it is correct, in that judicial judgement may not be surrendered to a machine. The comments that follow therefore go to institutional design, to the precision of a few key terms, and to one opportunity the draft has not yet taken. Each is framed as a specific amendment, with the present text and a proposed replacement, so that the Committee may consider it directly.

A single idea connects several of them. The draft already grades its safeguards by risk, requiring heightened scrutiny and human oversight for higher-risk uses (like in Regulations 7(3) and 12), yet it nowhere defines the risk tiers on which that gradation depends. A clear, defined risk classification, applied consistently to approval, audit, verification and disclosure, would lend the instrument much of its coherence: the lightest touch for routine administrative tools, and the firmest for anything affecting liberty. Several of the suggestions below build on that single spine.

We support the architecture and the guiding principles of the draft, in particular the principle of human primacy (Regulation 4), the placement of accountability on the responsible officer (Regulation 8), the express presumption in favour of responsible AI adoption (Regulation 16) and the “innovation over restraint” principle (Regulation 17), the permissible uses of AI for translation and accessibility (Regulations 19(c) and 19(g)) and the attention to linguistic diversity (Regulation 13), and the rule that a person who files AI-generated material bears responsibility for it and cannot use the AI character of the output as a defence (Regulation 43(6)). Our comments below are directed only at a small number of provisions where, in our practical experience, the drafting as it stands may produce effects we believe are unintended. 

Our comments are grouped as follows: Part A (matters we believe are of the highest importance), Part B (feasibility), Part C (clarifications and drafting points). Each comment cites the specific clause, states the concern, and proposes a concrete change.