The Supreme Court's 2026 Draft AI Rules: What Every Indian Advocate Must Know

Why the rules arrived now

The timing was not accidental. Over the preceding eighteen months, Indian courts had repeatedly caught fabricated, AI-generated case law slipping into filings and even into orders. In February 2026, a Bench led by Chief Justice Surya Kant openly expressed concern that lawyers were using AI to draft petitions containing non-existent precedents, calling the practice "uncalled for." The Bombay High Court had already imposed costs on a litigant for dumping AI-hallucinated citations into written submissions.

Against a backdrop of more than 5.4 crore pending cases nationwide, the judiciary wants the efficiency AI offers — but not at the cost of the authenticity that the doctrine of precedent depends on. The draft rules are the institutional answer to that tension.

What the draft regulations actually say

Three themes matter most for advocates.

First, a definition of the problem. The draft explicitly defines "AI hallucination" as the generation of fictitious or inaccurate legal content, including fabricated judicial precedents, statutory provisions, and quotations. Naming the risk in a regulatory instrument matters: it means courts now have a shared vocabulary for something they will increasingly scrutinise.

Second, AI cannot decide cases. The consistent thread across the draft — and across the High Court policies that preceded it — is that AI may assist with administrative and research tasks but must never substitute for judicial reasoning or the professional judgment of a lawyer. This mirrors how the Court uses its own tools: SUVAS for translation and SUPACE for research assistance, both strictly assistive.

Third, disclosure. The direction of travel is toward requiring that the use of AI be disclosed in appropriate circumstances. Even before any rule is finalised, the safe professional posture is transparency about how a document was produced and, crucially, verification of everything it contains.

What has not changed: your duty to the court

It is worth being clear about something the High Court policies of Kerala (July 2025), Gujarat (April 2026), and Punjab & Haryana make explicit: those policies bind judges and court staff, not practising advocates directly. But that does not let advocates off the hook.

An advocate's duty to verify authorities is not a new obligation invented for the AI era. It flows from long-standing professional responsibility. When the Delhi High Court saw a petition withdrawn in September 2025 after opposing counsel exposed its citations as fabricated — including invented paragraphs from a judgment that runs to only 27 paragraphs — the failure was not really about AI. It was about submitting authorities the advocate had never actually read. AI simply made that failure faster and more convincing.

How to adopt AI without regulatory risk

The draft rules are not a reason to avoid AI. They are a reason to use it correctly. Three practices keep you safe:

• Use grounded tools, not open chatbots for authority. A general-purpose chatbot predicts plausible text; it does not retrieve verified law. A legal platform built on your jurisdiction's actual judgment corpus — with every citation traceable to a real, linkable source — removes the single largest source of hallucination risk.
• Verify before you file. Treat every AI-surfaced citation as a lead, not a fact. Open the judgment. Confirm the paragraph. Read the ratio. This takes minutes and prevents career-damaging embarrassment.
• Keep a human in the loop for judgment. Let AI compress research and drafting time; keep the legal reasoning yours. That is exactly the boundary the Supreme Court is drawing for itself.

The takeaway

The 2026 draft regulations mark a maturing relationship between Indian law and artificial intelligence — one that welcomes the technology while insisting on accountability. Advocates who adopt AI thoughtfully will move faster, serve more clients, and spend their hours on argument rather than administration. Those who paste unverified output into filings will increasingly find courts prepared, and unforgiving.

The smartest response is not to wait for the final rules. It is to build a verification-first workflow now — one where AI does the heavy lifting on research and drafting, and every output is grounded in real, checkable Indian law. That is precisely the philosophy behind purpose-built legal platforms like LawWorld: speed you can defend in open court.

This article is for general information and does not constitute legal advice. The Draft Regulations for the Use of AI in Courts, 2026 remain under public consultation; advocates should follow the final notified rules and any Bar Council guidance.