Two Courts, Two Bets on the Same Problem
On June 1, 2026, New York's Unified Court System's Part 161 took effect, formalizing a wager: let lawyers use AI tools freely, don't make them say so, and rely on strict liability for fabricated citations to keep them honest. The rule states plainly that attorneys "should not be required, upon submitting papers, to disclose to the court" that they used AI — but any attorney who does use one must "carefully review the paper and independently ensure" it contains no fabricated cases or statutes, on pain of sanctions (NYSBA).
India's Supreme Court is drafting the opposite bet. Its AI Committee's draft Regulations for Use of Artificial Intelligence in Courts, 2026 — out for public comment until July 15, 2026 — includes a provision, Regulation 43, that would require anyone using AI in a legal filing to declare that use and show reasonable diligence in verifying it, a duty that appears to bind lawyers and litigants alike (Supreme Court Observer). New York trusts the threat of sanctions to deter sloppy use; India's draft wants a paper trail as well.
The Case That Forced India's Hand
This isn't abstract policy-making. On July 2, 2026, a Supreme Court bench of Justices P.S. Narasimha and Alok Aradhe set aside orders from the National Company Law Tribunal and the National Company Law Appellate Tribunal in an insolvency matter after finding that six precedents the NCLT relied on were compromised: three didn't exist at all, and three were real citations bolted onto invented paragraphs or wrong case titles. Crucially, the affidavit before the Court indicated counsel hadn't even cited these judgments — the tribunal appears to have surfaced the fake precedents through its own research, and the fabrication then slipped past an entire appellate tier undetected at the NCLAT (LiveLaw).
The Court did not mince words, comparing the spread of hallucinated case law to "the release of methyl isocyanate in the province of law and justice" — an unmistakable Bhopal reference — and calling reliance on fabricated precedent "misconduct on the part of an advocate" rather than a mere judicial slip. It directed the Bar Council of India to frame disciplinary norms for the practice. MediaNama has since catalogued ten separate instances of AI-hallucinated citations surfacing across Indian trial courts, tax tribunals, insolvency benches, and High Courts — evidence this is now a systemic pattern, not an isolated embarrassment (MediaNama).
The Case for Disclosure
India's instinct toward mandatory declaration deserves a fair hearing before it gets a pro-innovation rebuttal. Unlike most U.S. sanctions cases, where a lawyer's own hallucinated brief gets caught by an opposing counsel who checks the citations, the July 2 episode shows fabricated case law entering through a tribunal's own research and surviving appellate review entirely unchallenged. A disclosure requirement creates a flag for judges and opposing counsel to scrutinize AI-assisted portions of a filing more closely — exactly the kind of extra verification layer this case shows was otherwise absent. Given India's comparatively thin cadre of law clerks and the sheer docket volume moving through NCLTs and trial courts, a paper trail that helps triage where fabrication risk is highest is not an unreasonable ask.
Why Liability Alone Still Wins
But disclosure duties carry real costs that the insolvency ruling doesn't actually address. The fabrication in that case wasn't missed because nobody knew AI was involved — it was missed because nobody checked citations against a verified reporter, full stop. A disclosure line on a filing does nothing to fix that; the tribunal that generated the fake precedents through its own research would presumably have had to disclose its own AI use to itself. What actually would have caught it, as the iPleaders analysis on hallucinated case law in India argues, is the same discipline Part 161 mandates directly: independent verification against SCC Online, Manupatra, or the official reporters before anything reaches a bench (iPleaders).
Disclosure requirements also invite line-drawing disputes that add friction without adding safety — India's own drafters concede the current text is unclear on whether grammar-checking tools trigger the duty, or only generative drafting. New York's Part 161 sidesteps that entirely: it doesn't care what tool a lawyer used, only whether the citations in the final document are real, which is the only fact that actually matters to the court and to opposing counsel.
What India Should Take From New York
India's underlying instinct — AI assistance is fine, judicial authority is not delegable, and fabrication is misconduct — is sound and mirrors the core of Part 161. The Supreme Court's own draft regulations get the substantive line right: a hard ban on AI making judicial determinations, deciding bail, or profiling parties, with AI confined to transcription, translation, scheduling, and legal research (Supreme Court Observer). Where it should follow New York's lead is on disclosure: drop Regulation 43's declaration mandate and let strict liability for the underlying fabrication do the work, exactly as it did on July 2 when the Supreme Court didn't need a disclosure rule to find and punish the fake citations — it just needed to check them. A verification duty is enforceable and outcome-focused; a disclosure duty is a compliance checkbox that a hallucinating tribunal can satisfy on its way to citing case law that doesn't exist.