Artificial intelligence is already transforming the legal ecosystem, reshaping how lawyers conduct research, assisting courts in managing enormous backlogs, and, worryingly, generating fabricated or unreliable legal citations with disconcerting frequency. India’s Supreme Court has now decided it is time to get ahead of the curve. On June 3, 2026, it published the draft Regulations for Use of Artificial Intelligence in Courts, 2026, inviting comments from the public and stakeholders until June 20. The document is ambitious, detailed, and arrives not a moment too soon.
Why is it required?
- In recent times, AI use in legal procedure has repeatedly caused complications. In early 2026, the Supreme Court issued notices after a trial court relied on AI-generated, non-existent precedent. In February 2026, Justice B.V. Nagarathna encountered a reference to an entirely fictitious case in proceedings before her, and Chief Justice Surya Kant publicly slammed the growing practice of filing AI-drafted petitions without verification.
- These phantom precedents are what the draft calls “hallucinations” (defined in Regulation 3(z) as outputs that appear plausible or coherent but are factually incorrect, fabricated, misleading, or unsupported by verifiable data or source material). They pose a direct threat to the integrity of the justice system, and the draft regulations address this crisis head-on, while also going much further.
AI As Assistant, Not Judge
- The most fundamental principle running through the entire document is simple: humans stay in charge. Regulation 4 makes clear that AI systems shall function only in an assistive capacity and shall not supplant or compromise the independent exercise of judicial authority. The final authority on questions of law, fact, and justice remains exclusively with the judicial officers of the competent jurisdiction.
- Regulation 20(b) prohibits any judicial outcome from being reached through algorithmic decision-making alone or solely on the basis of AI-generated information, and Regulation 20(c) makes all AI output on adjudicative or sentencing questions advisory only, subject to independent judicial evaluation.
- Most significantly, the draft bans risk scoring entirely. No AI system can be used to assess flight risk, predict recidivism, evaluate bail eligibility, or determine the credibility of parties or witnesses. Risk-scoring tools have been widely criticized elsewhere for encoding racial and socioeconomic bias into sentencing decisions, and India’s draft firmly rejects that problem.
What AI Can Do in Courts
The growing adoption of these technologies is also influencing the work of artificial intelligence (AI) Lawyers in India, who are increasingly advising clients on the legal, ethical, and regulatory implications of AI-driven solutions.
- The regulations are not anti-technology. Regulation 16 creates a presumption in favour of responsible AI adoption, and Regulation 17 positions AI as a tool for meaningful innovation. The permitted uses under Regulation 19(1) are broad and practical: legal research and case summarisation, translation and transcription (subject to human verification), case management functions, accessibility tools, litigant-facing chatbots and conversational AI under supervision, and document authentication with mandatory human review.
- The regulations acknowledge that India’s courts are buried under staggering backlogs. As of March 2026, the Supreme Court alone had nearly 93,000 pending cases. AI, used wisely, is seen as a genuine tool to help clear this mountain.
What Has to Be Disclosed
- One of the most practically significant provisions is the mandatory disclosure rule under Regulation 43(3). If a lawyer or party uses AI to prepare any document, pleading, or piece of evidence, they must disclose this to the court at the time of submission by way of a duly executed declaration or certificate in a prescribed format (set out in Annexure I of the draft).
- Regulation 43(2) similarly requires courts to disclose when AI has materially assisted in case management or document analysis, while Regulation 43(5) mandates disclosure of any use of synthetic data or information in proceedings.
- Importantly, the draft eliminates the defence of blaming AI and filing parties remain fully liable for AI-generated submissions, directly addressing the risk of fabricated or misleading outputs without prohibiting the use of AI in legal practice altogether.
Institutional Architecture for AI Governance
- The draft creates a layered institutional structure to oversee AI in courts. At the top sits a permanent Apex Body at the Supreme Court level, comprising two Supreme Court judges, two Chief Justices of High Courts, two High Court judges, technology and cybersecurity experts, a finance specialist, a Ministry of Electronics and IT representative, and senior advocates. Below it, every High Court is required to constitute its own AI Committee of at least three judges, supported by an AI Secretariat staffed with judicial officers, data scientists, and technology experts and headed by an officer of District Judge rank.
- Regulation 32 proposes a Centre of Research and Excellence on AI (CoRE-AI) to conduct research, evaluate AI tools, and track global developments. All AI systems used in court must be registered, undergo annual technical and ethical audits, and be cleared through a mandatory Ethical Impact Assessment before deployment.
What AI Can Never Do
- The prohibitions in Regulation 20 are explicitly described as absolute and non-derogable, and cannot be subject to relaxation or modification by any authority. Beyond the bans on adjudication and risk scoring, Regulation 20 prohibits AI from being used to surveil court participants, profile or predict behaviour, train on personal data without prior approval, submit undisclosed AI-generated content as evidence, or compromise judicial confidentiality and independence.
- Under Regulation 21, every violation of these prohibitions must be reported immediately to the AI Secretariat, which places the matter before the AI Committee for remedial action.
Data Protection
- Chapter VII deals with data protection and cyber security. Regulation 48(1) restricts the transfer of sensitive judicial data to any external system, including private cloud infrastructure, without prior written approval.
- Regulation 46(9) ensures that AI tools built using court data remain court-owned, preventing private vendors from claiming intellectual property rights over tools built primarily using judicial data or public resources. It also mandates in-house audits, prohibiting the sharing of source code, algorithms, or training data with third parties.
Why This Matters Beyond India
India has one of the largest judicial systems in the world, and how it chooses to integrate AI may set a precedent for other common law jurisdictions grappling with the same questions. The draft regulations are fairly balanced: they refuse both the extremes of banning AI from courts entirely and of naively embracing it without guardrails. These developments are likely to create new opportunities for every technology law firm in India seeking to guide businesses, courts, and innovators through the evolving legal framework governing artificial intelligence. Regulation 57 even builds in a mechanism for continuous improvement, requiring periodic review of the regulations as AI technology evolves. With public comments open until June 20, 2026, the framework may still change, but even in draft form it reflects a proactive effort by the judiciary to address AI’s growing role.
Authors: Shantanu Mukherjee, Varun Alase
