Authored By: Mahewish Shaikh
- CASE CITATION AND BASIC INFORMATION
Case Name: Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr.
Citation: 2026 INSC 668
Civil Appeal No.: 11950 of 2025
Court: Supreme Court of India
Bench: Pamidighantam Sri Narasimha & Alok Aradhe, JJ.
Date: July 2, 2026
Nature: Civil appeal arising from insolvency proceedings under the Insolvency and Bankruptcy Code, 2016 (“IBC”).
Relevant Provisions: Insolvency and Bankruptcy Code, 2016, §§ 7, 14, Act No. 31 of 2016 (India).
Disposition: Appeal allowed; NCLT and NCLAT orders set aside; Section 7 application restored for fresh consideration.[1]
- INTRODUCTION
Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr., 2026 INSC 668, is a significant Supreme Court decision concerning the use of Artificial Intelligence (“AI”) in legal research and adjudication. Although the appeal originated from insolvency proceedings involving a corporate guarantee, the Supreme Court’s principal concern became the reliance of the National Company Law Tribunal (“NCLT”) upon judicial authorities that were either non-existent or attributed with paragraphs that could not be found in the judgments cited.[2]
The Court treated this not as an ordinary citation error but as a fundamental threat to the integrity of adjudication. It held that courts must maintain zero tolerance towards the production, citation, or reliance upon AI-generated precedents without verification.[3] At the same time, the Court clarified that legitimate use of AI is not prohibited; the objection is to presenting fabricated or hallucinated material as authentic legal authority.[4]
The judgment consequently establishes an important principle for AI-assisted legal practice: AI may facilitate research and efficiency, but the responsibility for verifying legal authorities remains with human advocates and adjudicators.
- FACTS OF THE CASE
Jammu & Kashmir Bank Ltd. had extended credit facilities to Pan India Utilities Distribution Company Ltd. (“PIUDCL”). Essel Infraprojects Ltd. (“EIL”) executed a corporate guarantee securing those facilities. Following financial difficulties and defaults by PIUDCL, its loan accounts were classified as non-performing assets.[5]
The Bank subsequently filed an application under Section 7 of the IBC before the NCLT, Mumbai, seeking initiation of the corporate insolvency resolution process against EIL as corporate guarantor. The NCLT admitted the application on August 28, 2024, finding debt and default and imposing the statutory moratorium.[6]
Pooja Ramesh Singh, a suspended director of EIL, challenged that order before the NCLAT. She argued, among other things, that EIL’s liabilities had been affected by restructuring within the group and that a renewed sanction letter dated November 18, 2017, did not refer to the corporate guarantee. The NCLAT rejected the challenge on September 11, 2025, relying, inter alia, upon Clause 8 of the guarantee deed, which preserved the guarantee notwithstanding absorption or amalgamation.[7]
Before the Supreme Court, however, counsel for the appellant questioned the authenticity of authorities relied upon by the NCLT. Upon independent examination, the Supreme Court found six problematic citations. Some referred to judgments that did not exist in the form presented, while another existing judgment had been accompanied by a paragraph that could not be traced to it.[8]
The respondent Bank subsequently clarified through an affidavit that its counsel had not cited those authorities before the NCLT and that the material appeared to have resulted from the Tribunal’s own research.[9] This discovery shifted the Court’s focus from the underlying guarantee dispute to the integrity of the adjudicatory process.
- LEGAL ISSUES
The principal issues before the Supreme Court were:
- Whether a judicial or quasi-judicial decision can stand when its reasoning relies upon fake, non-existent, or AI-hallucinated precedents.
- Whether advocates may place AI-generated authorities before a court without independently verifying them.
- Whether reliance by a judge or tribunal upon fabricated AI-generated material constitutes a serious lapse affecting the validity of the decision.
- What institutional measures are necessary to prevent AI-generated hallucinations from being presented as genuine legal precedent.[10]
The Supreme Court did not finally determine the substantive dispute concerning the corporate guarantee. Instead, it addressed the defect in the adjudicatory process and remitted the Section 7 proceedings for fresh consideration.[11]
- ARGUMENTS PRESENTED
5.1 Appellant’s Arguments
The appellant submitted that the authorities relied upon by the NCLT were fake, incorrectly attributed, or unsupported by the judgments cited. Counsel demonstrated that certain paragraphs reproduced in the NCLT’s reasoning could not be located in the corresponding judgments.[12]
The appellant also challenged EIL’s liability on substantive grounds, contending that restructuring arrangements had affected its obligations and that the renewed sanction letter did not preserve the corporate guarantee.[13]
5.2 Respondent’s Position
The respondent Bank defended the underlying insolvency proceedings and maintained its position regarding the debt, default, and corporate guarantee. However, its affidavit clarified that its advocates had not supplied the disputed authorities to the NCLT. According to the Bank, the material had emerged from the Tribunal’s own research.[14]
This clarification was significant because the case consequently demonstrated that the problem of AI-generated legal material is not confined to advocates; it may also arise through independent judicial or tribunal research.
- COURT’S REASONING AND ANALYSIS
6.1 AI as an Aid, Not a Substitute
The Court recognised AI’s increasing ability to perform both routine and intellectual tasks and acknowledged its potential to make professional work faster and more efficient.[15] Nevertheless, the Court cautioned against allowing technological capability to displace human reasoning and judgment.
It therefore stressed the necessity of maintaining “absolute and total control” over AI’s application, particularly in adjudication.[16] The Court’s approach was not anti-technology; rather, it required a human-controlled framework in which AI remains an assisting tool.
6.2 AI Hallucination and Legal Adjudication
The Court specifically addressed AI’s capacity to generate non-existent or fabricated material. In legal proceedings, the danger is particularly serious because an invented case may appear authoritative while having no legal existence.
The Court regarded such fabricated material as capable of contaminating the adjudicatory process and undermining the foundation of judicial determination.[17] The problem therefore extends beyond an inaccurate research result: a hallucinated authority creates a false source of law and may improperly influence judicial reasoning.
6.3 Zero-Tolerance Standard
The central holding appears in paragraph 7 of the judgment. The Supreme Court declared that courts must adopt a zero-tolerance approach towards producing, citing, or relying upon AI-generated precedents without verification. It characterised an advocate’s unverified citation of such material as misconduct and judicial reliance upon it as a serious lapse.[18]
The Court further held that a decision based upon fake or hallucinated material is “no decision in the eyes of the law,” regardless of whether the material had a direct or indirect influence upon the outcome.[19] Consequently, even limited entry of fabricated material into the decision-making process may require the decision to be set aside because the sanctity of adjudication has been compromised.
6.3 Responsibility of the Bar and Bench
The judgment imposes responsibility upon both sides of the adjudicatory system. Advocates must verify authorities before relying upon them. Judges and tribunals, while ordinarily entitled to rely upon counsel’s submissions, cannot permit fabricated authorities to become part of their reasoning.
The case is particularly important because the disputed material had apparently emerged from the Tribunal’s own research rather than from the respondent’s counsel.[20] Thus, verification is not merely an advocate’s obligation; it is an institutional requirement for reliable adjudication.
6.4 Legitimate AI Use Remains Permissible
The Court expressly distinguished between legitimate AI assistance and fabricated legal material. Its ruling does not prohibit the rightful use of AI in legal work.[21] The prohibited practice is presenting or relying upon hallucinated material as though it were an authentic judicial precedent.
6.5 Institutional Directions
Recognising that judicial pronouncement alone cannot resolve the broader problem, the Court directed the Bar Council of India to constitute a committee to consider the submission of fake or hallucinated material by members of the Bar. The committee was directed to formulate guiding principles and consider disciplinary consequences for violations.[22]
- JUDGMENT AND RATIO DECIDENDI
The Supreme Court set aside the NCLT order dated August 28, 2024, and the NCLAT judgment dated September 11, 2025. It restored the Section 7 application to the NCLT for fresh consideration, clarified that it had expressed no opinion on the merits, directed expeditious disposal, preferably within two weeks, and required the parties to maintain status quo until disposal.[23]
The ratio decidendi is that a judicial or quasi-judicial decision relying upon fake, non-existent, or AI-hallucinated legal precedent is legally unsustainable because such reliance compromises the integrity and sanctity of adjudication. Independent verification of legal authorities is therefore indispensable, and the tainted decision must be set aside even where the fabricated material did not demonstrably determine the ultimate result.[24]
The Court’s wider observations concerning technological dependence, human reasoning, and AI governance provide important jurisprudential guidance. However, the specific institutional direction to the Bar Council and the remand of the insolvency application should be distinguished from the core ratio.
- CRITICAL ANALYSIS AND CONTEMPORARY RELEVANCE
The decision’s greatest significance lies in recognising that AI hallucination creates a distinct category of legal risk. A conventional research mistake involves an incorrect understanding of an existing authority. An AI hallucination may instead create an authority that never existed. The latter is particularly dangerous because neither the court nor opposing counsel can meaningfully test its legal foundation.
The Court’s zero-tolerance approach is therefore institutionally defensible. Requiring proof that fabricated precedent actually altered the outcome would make judicial review of AI-generated material unnecessarily uncertain. By focusing on the integrity of the decision-making process itself, the Court establishes a clearer and more protective standard.[25]
For advocates, the practical lesson is direct: AI can be used to discover potentially relevant authorities, but every case name, citation, paragraph, proposition, and quotation must be checked against the original judgment or another authoritative source before being submitted to court.
For courts and tribunals, the decision highlights the need for reliable research infrastructure and human verification. AI-assisted judicial research should complement, not replace, judicial scrutiny.
The judgment is equally important for legal education. Law students and researchers increasingly use generative AI for case identification and drafting. The decision demonstrates why AI output should be treated as a research lead rather than as an independent source of law.
At the regulatory level, the direction to the Bar Council of India may lead to clearer professional standards concerning AI-assisted legal research, verification duties, disclosure practices, and disciplinary consequences. The Court itself recognised that broader AI governance cannot be achieved solely through judicial orders and requires appropriate policy and enforceable regulation.[26]
Nevertheless, the judgment leaves some practical questions unresolved. It does not establish a uniform verification protocol for every AI-assisted submission, nor does it comprehensively regulate the use of AI by courts, tribunals, and advocates. Effective implementation will also require access to authoritative databases, institutional training, and adequate research resources.
- CONCLUSION
Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr., 2026 INSC 668, represents an important judicial response to generative AI’s growing presence in the legal system. Although the appeal originated from an insolvency dispute involving a corporate guarantee, the Supreme Court ultimately focused on the more fundamental question of whether adjudication can remain legitimate when fabricated legal authorities enter judicial reasoning.
The Court answered firmly in the negative. It set aside the NCLT and NCLAT orders and restored the insolvency application for fresh determination, while declaring zero tolerance for unverified AI-generated precedents.[27] The judgment therefore establishes a clear principle for the future of AI-assisted legal practice: technology may assist legal reasoning, but it cannot replace human verification, judicial responsibility, or the authenticity of law.
- REFERENCE(S)
I.Primary Authority
- Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., Civil Appeal No. 11950 of 2025, 2026 INSC 668 (Supreme Court of India, 2 July 2026).
II.Statutory Provisions Referred To
- Insolvency and Bankruptcy Code, 2016 — Section 7.
- Insolvency and Bankruptcy Code, 2016 — Section 14.
III.Institutional/Official Material
- Supreme Court of India, Landmark Judgment Summary — Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., C.A. No. 11950/2025, identifying the subject [28][29]of the judgment as reliance upon AI-generated fake and hallucinated case law and the consequences for judicial decision-making.
- Supreme Court of India, Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., 2026 INSC 668, official judgment PDF.
[1] Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr., 2026 INSC 668, ¶¶ 18–21 (Supreme Court of India 2026).
[2] Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr., 2026 INSC 668, ¶¶ 1, 6–7 (Supreme Court of India 2026).
[3] Id. ¶ 7.
[4] Id. ¶ 7.
[5] Id. ¶ 10.
[6] Id. ¶¶ 11–12; Insolvency & Bankruptcy Code, 2016, No. 31 of 2016, §§ 7, 14 (India).
[7] Id. ¶ 13.
[8] Pooja Ramesh Singh, 2026 INSC 668, ¶¶ 14–15.
[9] Id. ¶ 16.
[10] Id. ¶¶ 7–9.
[11] Id. ¶¶ 18–20.
[12] Id. ¶¶ 14–15.
[13] Id. ¶ 12.
[14] Id. ¶ 16.
[15] Id. ¶ 2.
[16] Id. ¶ 5.
[17] Id. ¶ 6.
[18] Id. ¶ 7.
[19] Id. ¶ 7.
[20] Id. ¶ 16.
[21] Id. ¶ 7.
[22] Id. ¶ 9.
[23] Id. ¶¶ 18–20.
[24] Id. ¶ 17.
[25] Id. ¶ 17.
[26] Id. ¶¶ 5, 9.
[27] Id. ¶¶ 18–21.
