An Indian insolvency tribunal cited six fake judgments—three invented outright and three with passages that do not appear in the real cases cited—before the Supreme Court declared any ruling based on hallucinated material 'no decision at all.'
On July 2, 2026, India's Supreme Court voided a ruling from the country's insolvency tribunal that had been built on six AI-generated case citations. Three of those judgments were invented outright. Three more were real cases whose quoted passages do not appear in them. The bench disposed of the entire ruling in one sentence: a decision laced with hallucinated material is "no decision at all", and must be set aside even if "an iota of fake or hallucinated material enters the decision-making process."
The order does more than correct one bad ruling. It names a structural gap in how courts verify the authorities cited before them.
The ruling came from Justices P S Narasimha and Alok Aradhe, sitting on appeal from the National Company Law Tribunal (NCLT), the specialized court that handles corporate insolvency and bankruptcy cases in India. The NCLT had cited six judgments while rejecting objections raised by Essel Infraprojects Ltd against insolvency proceedings initiated against it as a corporate guarantor. The court had partly accepted a 2017 renewal-cum-reduction letter from Jammu and Kashmir Bank as evidence the corporate guarantee had been relinquished.
The six citations, however, were not what they appeared. ICICI Bank Ltd v Urban Infrastructure Real Estate Ltd (2019), V S Dempo & Co Ltd v Reliance Communications Ltd (2021), and Sarbjit Singh v Union Bank of India (2022) did not exist. Everest Kento Cylinders Ltd v Union of India (2015) and Canara Bank v N G Subbaraya Setty (2018) were genuine, but the passages quoted from them are not found in those judgments. The sixth citation, State Bank of India v M/s Shree Ram Urban Infrastructure Ltd, 2020 SCC OnLine SC 341, actually corresponds to a different case (M Subramaniam v S Janaki), and the quoted passage does not appear there either.
The bench did not treat this as a routine error. AI hallucinations, the court said, are "like the release of methyl isocyanate in the province of law and justice: invisible, insidious, and catastrophic", a reference to the 1984 Bhopal industrial disaster. The court's broader concern: AI could "infiltrate our intellectual work ethic and before long, render us dependent on its vast capabilities." The bench did distinguish AI's potential utility ("gratifying, even inspiring") from the unacceptable use of hallucinated material in adjudication.
The deeper story is the gap the ruling exposes. AI did not just produce wrong answers. It produced plausible, citeable wrong answers, formatted as case names, citations, and quoted passages, and those fabrications passed first-level judicial review in a live corporate-insolvency dispute. Common-law adjudication assumes the citations before a court are real. That assumption now has no procedural gate to enforce it.
Indian regulators are moving, though not directly at that gap. A Supreme Court panel has separately proposed a complete ban on using AI to decide verdicts or judge bail criteria, and the Bar Council has begun issuing guidelines on AI use in courts. Neither addresses the narrower problem the July 2 ruling surfaces: that lawyers can file fabricated citations with reasonable confidence no one at first instance will have time to verify each one.
The Essel matter now returns to the NCLT for a fresh hearing. The Supreme Court has not banned AI from Indian courtrooms. It has drawn a line: a citation that does not exist, or a quoted passage that does not appear in the cited case, voids the order it supports. The next test is whether lower courts treat that line as a rule or as a slogan.