Supreme Court declares zero tolerance for AI hallucinations in legal proceedings after NCLT case

Unchecked AI use in courts could be 'catastrophic': SC

The Supreme Court has issued a stern warning regarding the unchecked use of artificial intelligence in legal proceedings, deeming it potentially "catastrophic." Justices emphasized the grave threat posed by fabricated AI-generated legal precedents to the justice system's integrity. Consequently, the Bar Council of India is tasked with forming a committee to establish guidelines and disciplinary measures for AI in legal practice. The Supreme Court on Thursday emphasised the need for regulatory safeguards for the use of artificial intelligence (AI) in judicial proceedings, cautioning that its unregulated application could prove to be "catastrophic". "...if left unregulated, it may infiltrate our intellectual work ethic and, before long, render us dependent on its vast capabilities," a bench of justices PS Narasimha and Alok Aradhe said, stressing that the use of fabricated or non-existent AI-generated legal precedents could hurt the integrity of the justice system. Describing the use of fake and AI-generated hallucinated legal precedents as "invisible, insidious and catastrophic" to justice, the court likened it to "the release of methyl isocyanate (a highly toxic, flammable and volatile organic compound) in the field of law and justice". The court directed the Bar Council of India to set up a committee to examine the use of AI in legal practice and adjudication. It also asked the council to look into the issue of members of the bar submitting "fake and hallucinated material before the court as if it were legal precedent". The council "must take up this issue with utmost seriousness, deliberate earnestly, and prescribe guiding principles to prevent such occurrences, along with the disciplinary action that will follow a violation of the norms," it said. SC Sets Aside Essel Infraprojects Insolvency Order The SC also set aside orders passed by the Mumbai bench of the National Company Law Tribunal and the National Company Law Appellate Tribunal admitting Essel Infraprojects, a part of the Essel Group, into insolvency, for relying on non-existent, fake and hallucinated materials and verdicts generated through AI. It sent the matter back to the NCLT for fresh adjudication.

'Methyl Isocyanate of law': SC slams AI 'hallucinations', quashes NCLT verdict

The Supreme Court has declared zero tolerance for AI-generated fake or hallucinated judgments used as legal precedents. Setting aside an NCLT verdict, the apex court emphasized that such fabricated citations contaminate the justice system. The Bar Council of India is directed to form a committee to address this issue and prescribe disciplinary actions for lawyers submitting unverified AI material. The court stressed maintaining human oversight in AI-assisted adjudication. New Delhi: Red flagging the use of "non-existent, fake, and hallucinated" judgements generated by artificial intelligence (AI), the Supreme Court on Thursday set aside a verdict of the National Company Law Tribunal (NCLT), saying that courts must adopt zero tolerance against citing or using AI-generated precedents. The bench comprising Justices PS Narasimha and Alok Aradhe said, "The production of fake, non-existent, and hallucinated material and its utilisation as precedents in law, is like the release of methyl Isocyanate in the province of law and justice: invisible, insidious, and catastrophic by the time anyone notices. It not only contaminates but takes away the very lifeblood of judicial determination." Also Read: Lawyers welcome SC's AI rules against 'algorithmic justice' It set aside an NCLT judgement on Essel Infraprojects insolvency after finding that the tribunal had relied on non-existent, fake and hallucinated judgement precedents generated through AI tools. "It is necessary for Courts to adopt a zero-tolerance mode for producing, citing or using AI-generated precedents without verification. It is a misconduct on the part of an advocate to cite such judgments without verification," the verdict said. The bench said it was equally a "serious lapse" if a judge relies on such a fake or hallucinated AI-generated material as precedents in support of the determination. "We have no hesitation in declaring that such a decision is no decision in the eyes of the law, irrespective of whether such material had a direct or indirect bearing on the decision-making. Such decisions are to be set aside even if an iota of fake or hallucinated material enters the decision-making process, as it would violate the sanctity of adjudication," it said. The bench said it is absolutely necessary to maintain integrity in decision-making. "And we reiterate and declare zero tolerance for the Bar as well as the Bench to cite, refer to, or rely on such material. It is also clarified that our judgment shall have no bearing on the rightful use of AI, but on the presentation or reliance on fake or hallucinated material as if it were a court precedent," it said. The bench said mere declaration of prohibitory action is not sufficient and there must be a consequential action following accountability. "So far as the responsibility of the bar is concerned, we direct the Bar Council of India (BCI), being the apex statutory body, to constitute a committee and deliberate on this issue of members of the bar submitting such fake and hallucinated material before the Court as if they are precedents of law," it said. The verdict said the apex bar body must take up this issue with utmost seriousness, deliberate earnestly, and prescribe a guiding principle to prevent such occurrences, along with the disciplinary action that will follow a violation of the norms. The matter arose from an insolvency dispute involving Pooja Ramesh Singh and Jammu and Kashmir Bank Ltd and Essel Infraprojects Ltd. The appellant challenged an NCLT Mumbai order that had admitted a Section 7 insolvency application. The top court found that several "precedents" cited by the NCLT to justify its decision simply did not exist. These included fabricated case names and paragraphs wrongly attributed to genuine citations. For instance, the judgement cited ICICI Bank Ltd vs Urban Infrastructure Real Estate Ltd (2019) 16 SCC 528 and Sarbjit Singh vs Union Bank of India (2022) 7 SCC 464, both of which were found to be entirely non-existent citations. While respondent Jammu and Kashmir Bank filed an affidavit clarifying that their counsel had not cited these cases and that the NCLT had obtained them through its "own research," the apex court held that the source of the error did not mitigate the damage to the rule of law. "More than the inevitable consequence of setting aside such judgments, what is significant for our decision-making is our resolve to adopt AI technology in aid of adjudication, while at the same time asserting and declaring total and absolute control over adjudication, with a human in the loop at every stage," the bench observed.

10 cases that show Indian courts have an AI hallucination problem

The Supreme Court's latest judgment can be accessed here. The Supreme Court's (SC) July 2, 2026, ruling declaring zero tolerance for Artificial Intelligence (AI)-generated fake precedents was, in its own words, "yet again a case" of a tribunal relying on hallucinated material. Fabricated citations, judgments that do not exist, or real ones stuffed with invented paragraphs have surfaced across tribunals, trial courts, and High Courts (HCs) over the past year, with several cases clustered in the last seven months. Here are some cases that built up to the ruling: 1. Tribunal relied on six defective precedents, none cited by counsel (July 2026). A Bench of Justices PS Narasimha and Alok Aradhe set aside orders of the National Company Law Tribunal (NCLT) and the National Company Law Appellate Tribunal (NCLAT) in the Essel Infraprojects insolvency matter. Of the six judgments the NCLT relied on, three did not exist, and three real citations carried invented paragraphs or an incorrect case title. Jammu and Kashmir Bank's affidavit stated that its counsel had cited none of them, meaning the tribunal sourced the fake precedents through its own research, which then escaped scrutiny at the NCLAT. The SC called it misconduct for an advocate and a serious lapse for a judge, directed the Bar Council of India (BCI) to frame disciplinary norms, and likened hallucinated case law to "the release of methyl isocyanate in the province of law and justice." 2. SC declared fake AI citations "misconduct" and issued notices (February 2026). On February 27, 2026, the SC took cognisance of the Gummadi Usha Rani trial court order, declaring that a decision based on fake AI-generated judgments is "not an error" but "misconduct" with "legal consequences." It stayed reliance on the disputed report and issued notices to the Attorney General, Solicitor General and the BCI, appointing senior advocate Shyam Divan as amicus curiae. This was the direct precursor to the July ruling. 3. A fictitious "Mercy v. Mankind" alarmed the SC (February 2026). On February 17, 2026, a Bench of Chief Justice Surya Kant and Justices BV Nagarathna and Joymalya Bagchi flagged the growing use of AI to draft petitions while hearing a Public Interest Litigation (PIL) on political speeches. Justice Nagarathna said she had come across a non-existent case titled Mercy v. Mankind, and the CJI noted that in Justice Dipankar Datta's court "a series of such judgments were cited." 4. Bombay HC fined a litigant Rs 50,000 for "dumping" AI content (January 2026). In Deepak v. Heart & Soul Entertainment Ltd., the Bombay HC found that a litigant's submissions in an eviction dispute cited a judgment the court could not trace, Jyoti w/o Dinesh Tulsiani vs Elegant Associates. It flagged tell-tale signs of AI-generated output, including green tick-marks and repetitive phrasing, and imposed Rs 50,000 in costs. 5. Andhra Pradesh HC held fake citations do not always void an order (January 2026). A trial court in Vijayawada, in an August 2025 order, had dismissed objections to a commissioner's report by relying on four Supreme Court judgments that do not exist. When the losing party challenged the order, the Andhra Pradesh HC accepted that the citations were AI-generated but declined to set it aside in Gummadi Usha Rani vs Sure Mallikarjuna Rao. It reasoned that fake citations do not vitiate an order as long as the legal principle applied is correct, recording only a word of caution. It was this lenient approach that the SC later took up, holding instead that any decision tainted by hallucinated material is "no decision in the eyes of the law." 6. Bombay HC quashed a Rs 27.91 crore tax order built on non-existent precedents (October 2025). In KMG Wires Pvt Ltd v. National Faceless Assessment Centre, the Bombay HC found that the assessing officer had cited three non-existent decisions to justify a Rs 22.66 crore addition to a company's income. It quashed the order as a breach of natural justice, warning that quasi-judicial authorities cannot blindly rely on AI results. 7. A Delhi HC petition invented paragraphs from a landmark judgment (September 2025). In Greenopolis Welfare Association v. Narender Singh, a petition in a flat-possession dispute was withdrawn after opposing counsel exposed its citations as fabricated. It quoted paragraphs 73 and 74 of the landmark Raj Narain v. Indira Nehru Gandhi judgment, which runs to only 27 paragraphs. 8. Karnataka HC ordered a probe against a trial judge who cited fake SC rulings (March 2025). In a little-noticed order, Justice R Devdas of the Karnataka HC directed an inquiry and action against a City Civil Court judge in Bengaluru who had relied on two non-existent SC decisions to reject a return-of-plaint application in a commercial dispute. Senior counsel confirmed that neither side had cited those cases. It is one of the few instances where an HC moved against a judge, rather than just a lawyer, over hallucinated citations. 9. A tax tribunal ruled against a taxpayer using four non-existent judgments (December 2024). The Bengaluru bench of the Income Tax Appellate Tribunal (ITAT) recalled an order in Buckeye Trust v. PCIT, which had ruled in favour of the tax department in a Rs 669-crore dispute using three fake SC citations and one fake Madras HC citation. The department's representative had reportedly used ChatGPT, and the bench copied the citations without verification. 10. The Punjab and Haryana HC itself put a question to ChatGPT (March 2023). In Jaswinder Singh @ Jassi v. State of Punjab, Justice Anoop Chitkara asked ChatGPT about bail jurisprudence in cruelty cases and reproduced its answer in the order, clarifying it was only to present "a broader picture" and not an opinion on the merits. It was not a hallucination case, but it was the first instance of an Indian court openly incorporating generative AI into a judicial order.