Author: Shreya Singh, University of Allahabad
To the Point
On 2 July 2026, the Supreme Court of India delivered its first substantive ruling on artificial-intelligence-generated fake precedents. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (2026 INSC 668), the Court set aside orders of the National Company Law Tribunal and its appellate body after discovering that their reasoning rested on non-existent judgments and fabricated paragraphs falsely attributed to genuine case law. The Bench declared that any decision touched even marginally by hallucinated material is void and directed the Bar Council of India to frame disciplinary norms for advocates who cite unverified AI-generated authorities. The doctrine has since been applied to strike down a customs penalty exceeding four hundred crore rupees, signalling that its reach extends well beyond insolvency law into every corner of adjudication.
Use of Legal Jargon
A proper appreciation of the ruling requires familiarity with the following expressions, each of which carries precise doctrinal weight in this context:
• Stare decisis — the principle that courts are bound to follow established precedent.
• Ratio decidendi — the binding legal reasoning underlying a judgment, as distinct from obiter dicta.
• Res nullius — literally “belonging to no one”; used here to describe citations with no traceable judicial origin.
• Void ab initio — invalid from the very inception, as opposed to merely voidable.
• Audi alteram partem — the natural-justice principle requiring that the other side be heard.
• Per se — by itself, without reference to surrounding circumstances or degree.
• Professional misconduct — conduct falling below the standards expected of an advocate, actionable under Section 35 of the Advocates Act, 1961.
The Proof
The finding of fabrication rested on a clear evidentiary trail. First, several judgments relied upon by the Tribunal could not be traced in any law report, neutral citation database, or official court record — rendering them res nullius. Second, certain citations corresponded to genuine judgments, but the propositions extracted from them were entirely invented, meaning authentic vessels carried counterfeit cargo. Third, neither the advocates concerned nor the Tribunal had cross-checked the case name, citation, bench, and date before relying on the material — a minimal due-diligence step the Court held to be non-derogable. Significantly, the Bench held that the extent of fabrication was immaterial: even an indirect or marginal reliance on hallucinated material was sufficient proof of contamination, rendering the resultant order void ab initio.
Abstract
The rapid absorption of generative artificial intelligence into legal research has exposed Indian adjudication to a new hazard: reliance on non-existent, fabricated, and hallucinated case law. This article examines the Supreme Court’s ruling in Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (2026 INSC 668), where the Court set aside tribunal orders founded on AI-generated fictitious precedents and declared a zero-tolerance standard: a decision contaminated by even an “iota” of hallucinated authority is no decision in the eyes of law. The article traces the evidentiary basis of the finding, situates the ruling within a growing corpus of 2026 decisions on the subject, and evaluates the resulting obligations of verification cast upon the Bar and the Bench.
Case Laws
1. Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., 2026 INSC 668 — the locus classicus, holding that reliance on fake or hallucinated AI-generated precedent voids the resultant decision and amounts to professional misconduct on the part of the advocate concerned.
2. Vijay Ghanshyam Gadiya v. Union of India (Supreme Court, September 2026) — applied the same ratio to set aside a customs penalty exceeding ₹425 crore founded on non-existent case law.
3. Union of India v. [Anr.], 2026 INSC 947 — reiterated the zero-tolerance standard, extending the verification obligation to the Bench as well as the Bar.
4. Bombay High Court ruling (January 2026) — imposed exemplary costs of ₹50,000 for indiscriminate use of unverified AI-generated content in pleadings.
5. Andhra Pradesh High Court ruling (January 2026) — had earlier taken a more calibrated view, holding that a fabricated citation does not automatically void an order where the substantive reasoning survives independently, a position later superseded by the Supreme Court’s stricter standard.
Conclusion
The Pooja Ramesh Singh ruling marks a doctrinal watershed in the relationship between artificial intelligence and Indian adjudication. By distinguishing legitimate technological assistance from unverified reliance, the Court has charted a sound middle path that neither rejects AI outright nor permits its unchecked use. The responsibility now lies with the Bar Council of India to translate this mandate into concrete disciplinary rules, and with individual practitioners to treat independent verification as a non-negotiable professional discipline in an era where courts themselves increasingly deploy AI tools for research and translation.
FAQs
Q1. What did the Supreme Court hold in Pooja Ramesh Singh? That any decision resting, even partly, on fake or AI-hallucinated precedents must be set aside as void.
Q2. Is citing a hallucinated case professional misconduct?Yes, the Court held it attracts liability under the Advocates Act, 1961, regardless of the technology used.
Q3. Has the Court banned the use of AI in legal research?No. It permitted AI assistance but mandated independent verification before reliance.
Q4. Does the extent of fabrication matter? No, the Court rejected a proportionality test, holding even an “iota” of fabrication sufficient to void an order.
Q5. What must a lawyer verify before citing a precedent?The case name, citation, deciding bench, date, and whether the cited proposition genuinely appears in that judgment.
