POOJA RAMESH SINGH v. JAMMU AND KASHMIR BANK LTD. & ANR.

Case Name: POOJA RAMESH SINGH v. JAMMU AND KASHMIR BANK LTD. & ANR.

Citation: 2026 INSC 668; Civil Appeal No. 11950 of 2025.

Court: Supreme Court of India

Date of Decision: 2 July 2026.

Bench: Justice Pamidighantam Sri Narasimha & Justice Alok Aradhe.

Introduction:

Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. & Anr. happens to be one of the earliest judgments of Supreme Court of India to address the question of consequences of relying on judicial authorities which do not exist. It emerged out of an ordinary corporate insolvency case under Section 7 of the Insolvency & Bankruptcy Code, 2016, but it became a matter of great importance when Supreme Court revealed that order of National Company Law Tribunal was based, in part on judicial precedents created by AI & were never actually decided. The Bench took this opportunity to formulate a principle for AI-generated documents in adjudication & further directed the Bar Council of India to look into issue of professional conduct. Hence, the verdict holds relevance for bankruptcy law, reliability of evidence & ethics of use of technology in legal profession[1]. Therefore, the ruling relates to issues concerning insolvency legal system, evidential reliability as well as ethics associated with technology-ridden practice & forms a part of similar controversies in USA & UK.

Facts of the case:

On December 17, 2013, Jammu & Kashmir Bank Ltd. sanctioned a long-term working capital loan of 200 crore rupees in favour of Pan India Utilities Distribution Company Ltd. To ensure repayment, Essel Infraprojects Ltd., a group company, gave a corporate guarantee to the bank and created a mortgage over approximately 196 acres of land at Gorai Village, Borivali (West). The loan agreement was executed on December 27, 2013 & was renewed through a renewal-cum-reduction letter on November 18, 2017.[2]

Following this, PIUDCL faced financial difficulties & failed to meet its financial commitments. As a result, the Bank identified that this loan account had become a non-performing asset. The Bank, thereby, used the corporate guarantee & sent a petition against EIL to NCLT, Mumbai Bench, under Section 7 of the IBC, claiming that a debt of around Rs. 87.43 crore was outstanding & due to be settled by EIL as corporate guarantor.

By its order dated 28 August 2024, the NCLT allowed the petition under Section 7, appointed an Interim Resolution Professional, suspended Board of Directors of EIL (including the appellant, Pooja Ramesh Singh, an acting director) & provided for a suspension under Section 14 of IBC. The appellant was unhappy with this order & went on to appeal before the National Company Law Appellate Tribunal. The Tribunal initially stayed all proceedings but finally dismissed the appeal on 11 September 2025, making reference to legal precedents cited in NCLT’s order.[3] The incident was only uncovered when the Supreme Court took up the civil appeal & eventually the bench learned that most of the precedents which were cited by the NCLT & also verified by the NCLAT without conducting independent examination, had all been fabricated & some of the texts manipulated with some material even referenced from genuine citations. The Bank in its affidavit stated that the given precedents were not produced before the court by their attorney but somehow found their way into deliberation process based on research of the tribunal.[4]

Legal Issues:

The Supreme Court framed, expressly or by necessary implication, the following questions:

1. Whether an order of a tribunal or appellate tribunal that relies, even in part, on non-existent, AI-hallucinated case law can be sustained in law.

2. Whether the citation of unverified, AI-generated precedents by an advocate before a court or tribunal amounts to professional misconduct.

3. What duty of verification rests on adjudicating authorities, including appellate tribunals, before relying on precedents placed before them?

4. Whether, on the merits, the corporate guarantee executed by Essel Infraprojects Ltd. continued to subsist so as to sustain the Section 7 application, a question the Court left open for fresh determination.

Arguments Presented:

Appellant’s Arguments:

The appellant argued that the corporate guarantee provided by EIL had either been cancelled or transferred to another company because of a demerger process & thus the demanded money from EIL could not be collected. More importantly, when the false nature of previous cases became known, the appellant stated that the order of the NCLT & the ruling of NCLAT were flawed from the very beginning because any decision based on non-existing facts cannot be considered as legally valid.

Respondent’s Arguments:

The Bank asserted that the fact of the existence of debt & default was independently proved from the information already available, without taking into account the disputed references. The Bank distanced itself from the false precedents, even filing an affidavit clarifying that its lawyer did not refer to them and that they appeared to originate from the tribunal’s own, probably AI-assisted research. While being in favour of the rectification of the record, the Bank insisted on allowing the insolvency proceeding to go on without being terminated completely.[5]

Court’s Reasoning & Analysis:

The Division Bench concluded that this was not an isolated instance, but rather part of a larger pattern wherein tribunals have relied unknowingly on fictions or non-existent material created by artificial intelligence & accepted it as credible authority. A ruling based on such information will therefore be flawed. Whenever the reasoning in the judgment is based on fictitious citations, the judgment itself is rendered meaningless as it reflects a violation of the mandate of the law rather than a use of the adjudicatory process.

The Bench was equally disturbed by the fact that such counterfeit judgments had not only escaped the attention of NCLT, but also of the NCLAT, the very first appellate forum set up by legislation for this purpose. Courts & tribunals cannot but rely on materials presented to them by advocates, but this trust cannot be unconditional. Consequently, the Court urged courts & tribunals to enforce “complete & total control” over utilization of AI in judicial process, which implies conscious choices as to when & how to use this tool & not treating the result of its functioning as unquestionable.[6]

On accountability, the court adopted a zero-tolerance stance: using AI-generated judgments without validating their authenticity is considered as a form of misconduct by a person in legal profession & so is trusting its findings without validating their authenticity by a judge or tribunal member. According to court’s ruling, in terms of law & order based on a false or imaginary finding is not a ruling regardless of its significance for the case it is enough for the ruling to contain only some fragments of the falsified information for violation of the whole process of finding of truth. The court also instructed Bar Council of India to consider the misuse of AI technology by lawyers & develop reasonable regulations.

The Court framed its judgement using a comparative perspective, invoking the US judgement in Mata v. Avianca, Inc., whereby a lawyer in New York was punished for submitting a brief that included citations produced by AI.[7] The Court also took a reference to the UK case of R (Ayinde) v. London Borough of Haringey, the first of its kind in the UK, where AI was misused by a lawyer during the hearings. The Court went beyond that as it imposed consequences on the tribunal beyond individual sanctions.[8]

Judgement & Ratio Decidendi:

The Supreme Court removed the orders of both the NCLT from 28 August 2024 & of NCLAT from 11 September 2025. It restored to the NCLT the Section 7 application of Jammu and Kashmir Bank Ltd. on its original number with the direction to dispose the application de novo and in accordance with law, without considering any of the evidence based on a fabricated document. The Court also directed the Bar Council of India to look into the issue of AI-generated bogus precedents and to take such regulatory measures as it deemed appropriate.

The ratio decidendi that comes out is that any judicial or quasi-judicial order with even a minor element of a non-existent or AI-generated precedent becomes void & unmaintainable regardless of how much impact that evidence has had on the outcome; the obligation to ensure the correctness of the jurisprudence cited lies with both the lawyer citing the authority and the judicial authority relying on it, and the fact that such a citation has already been accepted by a lower authority does not relieve the appellate body of its obligation to inspect the citation.[9]

Critical Analysis:

1. Significance of the decision:

This verdict is noteworthy for being one of the very first official Indian judgments to interpret AI hallucination as more than just a minor irregularity but, on the contrary, as a serious flaw in itself that can tarnish the integrity of the entire adjudication process. By voicing its interest in the process of appellate review and by ordering the Bar Council of India to tackle that concern, the Court made it clear that the problem is of systemic nature and is not confined to any offence of a particular litigant but, rather, is highly relevant for a nascent branch of comparative jurisprudence dealing with technology and the court process.[10]

2. Implications & Impact:

As per the decision, the advocate has to responsibly ensure that each citation has to be verified and compared with trustworthy sources, such as official court websites, Indian Kanoon, or reputable law reports before usage. Any use of AI-generated data without verification opens them to the risk of being charged with professional misconduct. This ruling provides impetus for such high-volume forums like NCLT and NCLAT to prepare internal verification strategies and train officials in spotting hallucination detection. For example, the innocent party like the bank in this case, the end result is delays as the creditor must wait for the upcoming deliberations without doing anything wrong.[11]

3. Critical Evaluation:

The key advantage of the ruling is that it is straightforward and clear: instead of measuring consequences based on the degree of the impact of an artificial citation, the Court has established an unequivocal law that says any artificial evidence makes decisions invalid, which has a strong deterrent effect and helps keep the integrity of the judiciary intact. It also correctly points out that the responsibility is laid on both the bench and the lawyers rather than just the lawyers.

Nevertheless, the ruling leaves a number of questions open. It does not provide a criterion for differentiating legitimate AI-aided research from the unacceptable use of unverified results of AI and it does not suggest a way to discover artificial evidence before making a ruling. The rule stating that even a minor artificial evidence negates a decision regardless of its actual impact is clear from a legal point of view but may be considered impractical from the perspective of judicial economy, as it establishes a requirement for complete overturning of the decision even if a decision based on independent findings regarding the debt and default could have been made.[12]

Conclusion:

Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr. is expected to be remembered more for the precedent it sets at the intersection of technology and adjudication, rather than the contribution it has made to the field of law on insolvency, as it clearly states that AI can help in legal research, but can never replace verification of facts and accountability required in human judgment during the judicial process. The court has set aside two invalid decisions and has sent the matter back for fresh hearing and has directed the Bar Council of India to consider ethical implications of AI usage, which amounts to the court trying to put some restrictions on the use of AI in Indian courts and tribunals. Whether the corporate guarantee provided by Essel Infraprojects Ltd. stands the test of time as per the insolvency claim of the bank remains to be seen at the NCLT, but it is certain that no future decision can be based on non-existent precedents.

Reference(S):

[1] Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., 2026 INSC 668 (India).

[2] Id.

[3] Supra note, 1.

[4] Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., (2025) ibclaw.in 738 NCLAT (NCLAT 2025).

[5] Pooja Ramesh Singh v. J&K Bank (2026): SC Tears Apart NCLT Order That Relied on AI-Hallucinated Judgments, Aashayein Judiciary (2026).

[6] V. Venkatesan, Fake AI Cases ‘Entered’ an NCLT Insolvency Order. The Supreme Court Quashed It. But Who Is Accountable?, The Wire (2026)

[7] Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).

[8] Ayinde v London Borough of Haringey, and Al-Haroun [2025] EWHC 1383.

[9] Utkarsh Suman, The Limits to Delegation: Analyzing Pooja Ramesh Singh Vs Jammu and Kashmir Bank Ltd, Indian Journal of Law and Legal Research (2026).

[10] Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., Comp. App. (AT) (Ins) No. 1808 of 2024, CourtMesh (2025),

[11] Id.

[12] Lawbeat News Desk & Salil Tiwari, ‘AI Should Aid, Not Replace Human Reasoning’: Supreme Court on Fake Case Law, LawBeat (2026)