Inside the Supreme Court's Draft AI Regulations for Courts, 2026

On 3 June 2026, the Supreme Court of India's Artificial Intelligence Committee - headed by Justice P.S. Narasimha - published a draft titled the Regulations for Use of Artificial Intelligence in Courts, 2026, and invited public comment through 20 June 2026. It's the first structured attempt by India's judiciary to set formal terms for how courts and associated bodies may adopt AI across judicial and administrative work.

The draft's organizing idea is human primacy: AI may play a strictly assistive role, and the power to decide stays with judges alone. Permitted uses include legal research, summarization, translation, transcription, scheduling, litigant-assistance chatbots, and court analytics.

Explicitly prohibited: AI-only decision-making, AI-based risk scoring for bail or credibility assessments, predictive profiling of parties or witnesses, surveillance of judges or lawyers, and undisclosed AI-generated evidence. That prohibition list is nearly a checklist of the AI-in-justice-system risks that have drawn the most criticism internationally - the draft addresses them directly rather than leaving them to case-by-case interpretation.

The most operationally significant piece for practicing advocates is draft Regulation 43(3): any party or legal representative using AI in the preparation of pleadings, documents, submissions, or evidence would be required to disclose that AI-assisted origin at the time of filing.

This is a proposal, not yet a binding rule - but firms and advocates already using AI drafting or research tools should treat it as a strong signal of where the rule is heading, and start thinking now about how they'd document AI involvement in a filing if disclosure becomes mandatory.

The Supreme Court's draft follows the Kerala High Court's July 2025 policy on the use of AI tools in the district judiciary, which set an earlier example of court-level AI governance built around the same idea - AI as an assistive tool, not a substitute for legal reasoning. Together, these signal that AI governance in the Indian judiciary is moving from individual-court experimentation toward a more national framework.

This is a draft. The public comment window closed 20 June 2026, and the final form of any adopted regulation may differ from what was published. Readers who need the current, binding status should check the Supreme Court's own notices and its e-Committee website directly rather than relying on secondary summaries, including this one.

The draft's central distinction - AI assists, humans decide, and outputs need to be traceable to a source rather than trusted on fluency alone - is the same principle behind LegalDreams' own 'source-grounded, not just fluent' approach to AI. A judiciary-level framework converging on the same standard is a good sign for where legal AI tooling generally needs to head, not a constraint specific to any one vendor.