Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd.: Zero Tolerance for AI-Hallucinated Precedents and the Duty of Human Verification

Name: Mahima Mittal
College: Asian Law College, CCSU

To the Point

The Supreme Court’s decision in Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., 2026 INSC 668, arose from an appeal against the National Company Law Appellate Tribunal’s (NCLAT) decision affirming the National Company Law Tribunal’s (NCLT) admission of insolvency proceedings under Section 7 of the Insolvency and Bankruptcy Code, 2016 (IBC).

The appellant, a suspended director of the corporate guarantor, challenged the insolvency proceedings and disputed continuing liability under the corporate guarantee. The proceedings, however, assumed a much wider significance when the Supreme Court independently examined the authorities relied upon by the NCLT and discovered that several were either non-existent, incorrectly cited, or contained fabricated paragraphs attributed to genuine judgments.

The case consequently raised a question extending beyond insolvency law: what happens when adjudication is influenced by legal authorities that do not exist? The Supreme Court responded with a stringent approach, declaring zero tolerance for the production, citation or reliance upon AI-generated fake or hallucinated precedents without verification. At the same time, the Court clarified that its judgment was not intended to prohibit the rightful use of Artificial Intelligence (AI).

Use of Legal Jargon / Legal Language

The Bench comprising Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe examined the matter against the principles of judicial integrity, professional responsibility, application of judicial mind, authenticity of precedent and sanctity of adjudication.

Pan India Utilities Distribution Company Ltd., the principal borrower, had obtained credit facilities from Jammu and Kashmir Bank Ltd. Upon default, the corporate guarantee furnished by Essel Infraprojects Ltd. was invoked. The Bank thereafter initiated proceedings under Section 7 of the IBC against the corporate guarantor. On 28 August 2024, the NCLT admitted the application, appointed an Interim Resolution Professional and imposed a moratorium under Section 14 of the IBC.

The appellant challenged the admission before the NCLAT, contending, inter alia, that the corporate guarantee had ceased to operate following a restructuring involving demerger and amalgamation. The NCLAT rejected the challenge on 11 September 2025, relying upon the guarantee deed and authorities referred to in the NCLT’s reasoning.

The Supreme Court thereafter examined the authorities relied upon by the NCLT. Six problematic references were identified. State Bank of India v. M/s Shree Ram Urban Infrastructure Ltd. contained an incorrect citation to an existing judgment and a non-existent paragraph; Everest Kento Cylinders Ltd. v. Union of India had a correct citation but a non-existent paragraph; ICICI Bank Ltd. v. Urban Infrastructure Real Estate Ltd. and V.S. Dempo & Co. Ltd. v. Reliance Communications Ltd. were found to be non-existent citations; Canara Bank v. N.G. Subbaraya Setty & Anr. had a correct citation but a non-existent paragraph; and Sarbjit Singh v. Union Bank of India was also found to be non-existent.

The Bank informed the Supreme Court that these authorities had not been cited by its counsel and that the so-called precedents had been obtained through the adjudicating authority’s own research. This distinction became significant because the issue was not merely an advocate placing unreliable material before the Court; fabricated legal material had entered the adjudicatory reasoning itself.

The judgment therefore establishes a critical distinction: AI may assist legal research, but an AI-generated output cannot acquire the status of legal authority without independent verification. Human responsibility remains attached to the advocate, judge or adjudicatory authority relying upon the material.

The Proof

The judgment demonstrates several forms of AI-related unreliability in legal research: an incorrect title attached to a genuine judgment, fabricated paragraphs attributed to authentic decisions, and entirely non-existent authorities.

This is particularly serious because an AI hallucination may appear legally authentic. A fabricated authority can contain a plausible case title, citation, court, date, paragraph number and legal proposition. Consequently, verification requires more than reproducing a citation; the underlying judgment must actually be located and examined.

The Supreme Court’s intervention is significant because the problem was not confined to the conduct of an advocate. The fabricated authorities had entered the reasoning of the adjudicatory process and were relied upon by the NCLT and subsequently considered by the NCLAT.

The Court consequently treated the issue as one affecting the sanctity of adjudication itself. It held that a decision based upon fake, non-existent or hallucinated AI material cannot be sustained merely because the final conclusion might independently appear correct. The presence of such material in the decision-making process undermines the authenticity of the judicial reasoning upon which the decision rests.

Abstract

Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr.is a significant Supreme Court decision at the intersection of insolvency law, Artificial Intelligence and judicial responsibility. Although the dispute originated in a Section 7 IBC proceeding, the case developed into a broader examination of fabricated legal authorities after the Supreme Court discovered that several citations relied upon by the NCLT were either non-existent or contained hallucinated passages.

The Court adopted a zero-tolerance approach towards the production, citation and reliance upon fake or hallucinated precedents without verification. It treated such reliance as fundamentally inconsistent with the integrity of adjudication and professional responsibility. The Court further emphasised that a judicial or adjudicatory determination affected by such material cannot be preserved merely because the conclusion might otherwise appear sustainable.

Importantly, the judgment does not amount to a blanket prohibition on AI. The concern is not technological assistance itself but the substitution of unverified machine-generated material for authentic legal authority. The judgment therefore raises a broader principle: AI may assist legal reasoning, but responsibility for verifying, evaluating and applying the law remains human.

Case Laws

1. Gummadi Usha Rani & Anr. v. Sure Mallikarjuna Rao & Anr.

CRP No. 2487 of 2025, Andhra Pradesh High Court, decided 21 January 2026; SLP(C) No. 7575 of 2026, Supreme Court

The Andhra Pradesh High Court considered a matter involving non-existent authorities and examined the responsibility of parties and counsel to verify legal material. The High Court discussed the serious implications of placing before a court authorities that do not exist or do not support the propositions for which they are cited.

The matter subsequently reached the Supreme Court in SLP(C) No. 7575 of 2026. On 27 February 2026, the same Bench comprising Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe expressed considerable institutional concern regarding the process of adjudication and the use of AI-generated, non-existing or fake judgments. The Supreme Court treated the issue as having a direct bearing upon the integrity of the adjudicatory process.

Thus, Gummadi Usha Rani should not simply be described as having been “overruled” by Pooja Ramesh Singh. Rather, the proceedings demonstrate the development of judicial thinking concerning AI-generated authorities, with Pooja Ramesh Singh subsequently articulating an express zero-tolerance approach.

2. KMG Wires Private Limited v. National Faceless Assessment Centre & Ors.

W.P. (L) No. 24366 of 2025, Bombay High Court, decided 6 October 2025

The Bombay High Court dealt with an assessment order containing, among other defects, reliance upon three judicial decisions that were found to be non-existent. The Court specifically observed that quasi-judicial authorities cannot blindly rely upon results generated by AI tools and must independently verify such material. The assessment order and consequential proceedings were quashed and the matter was remanded for fresh consideration.

The case is relevant because it demonstrates that the duty of verification applies not merely to traditional courts but also to administrative and quasi-judicial authorities.

3. Mr. Deepak s/o Shivkumar Bahry v. Heart & Soul Entertainment Ltd. & Anr.

W.P. No. 8390 of 2009, 2026:BHC-AS:828, Bombay High Court, decided 7 January 2026

The Bombay High Court encountered written submissions containing features that strongly indicated the use of an AI tool and a cited case that neither the Court nor its law clerks could trace. The Court stated that the use of AI in aid of research is welcome, but emphasised the corresponding responsibility of parties and advocates to cross-verify references and ensure that generated material is relevant, genuine and in existence. Costs of ₹50,000 were imposed.

The decision directly complements Pooja Ramesh Singh by demonstrating that AI-assisted research does not diminish the duty of professional verification.

4. Buckeye Trust v. Principal Commissioner of Income Tax

ITA No. 1051/Bang/2024, Income Tax Appellate Tribunal, Bengaluru

Buckeye Trust is relevant only as an illustrative tribunal-level development concerning the use or suspected use of AI-generated material in adjudicatory proceedings. It should not be treated as a binding authority equivalent to a Supreme Court or High Court precedent.

Its relevance lies in demonstrating the practical difficulty created when adjudicatory documents contain legal material whose authenticity is uncertain. The case therefore supports the broader proposition that machine-generated legal material requires human verification before it is incorporated into judicial reasoning.

Conclusion

Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr.is significant not because it declares AI incompatible with the legal system, but because it establishes a boundary between technological assistance and technological substitution.

The Supreme Court’s reasoning rests upon a fundamental proposition: the ultimate responsibility for legal reasoning remains with the human advocate, judge or adjudicating authority who relies upon the material. An AI system may assist in locating, organising or processing information, but it cannot authenticate its own output, confer precedential authority upon an invented judgment, or replace judicial application of mind.

The Court set aside the NCLT and NCLAT orders after finding that fabricated and hallucinated authorities had entered the adjudicatory process. It restored the Section 7 application to its original number and directed the NCLT to proceed with the matter in accordance with law, leaving the merits open. The Court also directed the Bar Council of India to examine the issue and consider appropriate disciplinary norms concerning the citation of unverified AI-generated material.

The judgment nevertheless leaves important questions for future legal policy. Should professional consequences distinguish deliberate fabrication from negligent failure to verify an AI-generated citation? Should advocates disclose material reliance upon generative AI? Should courts and tribunals adopt formal verification protocols? These questions should not be presented as existing law; they represent areas for further professional and institutional development.

The central lesson, therefore, is not that technology must be excluded from legal practice. Rather, technological efficiency cannot displace the duty of verification, professional accountability and judicial application of mind. AI may assist adjudication; responsibility for adjudication, however, remains human.

FAQs

1. Did the Supreme Court ban AI in legal practice?

No. The judgment does not impose a blanket prohibition on Artificial Intelligence. The Court expressly distinguished legitimate use of AI from presenting or relying upon fabricated or hallucinated material as genuine legal authority.

2. What is an AI hallucination in legal research?

An AI hallucination occurs when a generative AI system produces information that appears plausible but is false or unsupported. In legal research, this may include invented cases, incorrect citations, fabricated quotations or non-existent paragraphs.

3. Why is a fabricated precedent more serious than an ordinary factual error?

A fabricated precedent carries the appearance of judicial authority even though no such authority exists. If relied upon in adjudication, it can distort legal reasoning and undermine the integrity and legitimacy of the decision-making process.

4. Who is responsible for verifying AI-generated legal material?

The human user who relies upon the material remains responsible for verification. An advocate cannot shift professional responsibility to an AI system merely because the system generated the citation; similarly, a court or tribunal must independently satisfy itself regarding the authenticity and relevance of authorities forming part of its reasoning.

5. What is the central principle of Pooja Ramesh Singh?

The central principle is that AI may assist legal work, but fabricated or hallucinated authorities cannot substitute authentic precedent, and human verification and judicial application of mind remain indispensable to adjudication.

 

 

References

1. Supreme Court of India — Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., Civil Appeal No. 11950 of 2025, 2026 INSC 668 (2 July 2026).
Judgment / Full Text

2. Supreme Court of India — Gummadi Usha Rani & Anr. v. Sure Mallikarjuna Rao & Anr., SLP(C) No. 7575 of 2026, order dated 27 February 2026.
Supreme Court Order

3. Andhra Pradesh High Court — Gummadi Usha Rani & Anr. v. Sure Mallikarjuna Rao & Anr., CRP No. 2487 of 2025, decided 21 January 2026.
Judgment / Full Text

4. Bombay High Court — KMG Wires Private Limited v. National Faceless Assessment Centre & Ors., W.P. (L) No. 24366 of 2025, decided 6 October 2025.
Case Report / Judgment Details

5. Bombay High Court — Mr. Deepak s/o Shivkumar Bahry v. Heart & Soul Entertainment Ltd. & Anr., W.P. No. 8390 of 2009, 2026:BHC-AS:828, decided 7 January 2026.
Judgment / Full Text

6. Income Tax Appellate Tribunal, Bengaluru — Buckeye Trust v. Principal Commissioner of Income Tax, ITA No. 1051/Bang/2024.
ITAT / Case Material