Abstract
The increasing use of artificial intelligence ("AI") in legal research and judicial decision-making has introduced a critical challenge for the administration of justice: the generation and reliance upon fictitious judicial precedents. In Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., the Supreme Court of India examined the consequences of reliance upon AI-generated case citations and judicial extracts that were either non-existent or incorrectly attributed to genuine judgments.1 The Court set aside the decisions of the National Company Law Tribunal ("NCLT") and the National Company Law Appellate Tribunal ("NCLAT"), emphasising that adjudication must remain grounded in authentic legal material and subject to human judgment. This article examines the factual background, ratio decidendi, professional implications and practical consequences of the decision for advocates, courts, tribunals and legal institutions. It argues that responsible AI-assisted legal practice requires independent verification, professional accountability and institutional safeguards against the uncritical adoption of machine-generated legal content.
Keywords: Artificial Intelligence, AI Hallucinations, Legal Research, Professional Ethics, Judicial Integrity, Supreme Court of India, Insolvency and Bankruptcy Code, 2016.
I. Introduction
Artificial intelligence is increasingly transforming the manner in which legal professionals undertake research, draft pleadings, analyse judicial decisions and prepare submissions. Generative AI systems can identify legal issues, summarise lengthy judgments and assist in organising complex arguments. However, these capabilities are accompanied by a significant limitation: AI systems may produce plausible but factually incorrect information, including fictitious judicial decisions, inaccurate citations and passages falsely attributed to existing judgments.
In legal practice, such errors are particularly consequential. Judicial precedents are not merely informational references; they form an essential component of legal reasoning and adjudication. A fabricated authority may misrepresent the law, distort a party's submissions and undermine the reliability of a judicial determination.
This concern received authoritative judicial consideration in Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., a judgment delivered by the Supreme Court of India on 2 July 2026.2 The Court considered the consequences of reliance upon six purported judicial authorities that were found to be non-existent, incorrectly cited or associated with paragraphs that could not be traced to the judgments concerned.
The decision is significant for two reasons. First, it establishes a clear judicial position on the consequences of introducing fictitious legal material into adjudication. Second, it addresses the professional and institutional responsibilities arising from the increasing use of AI in legal research and decision-making.
This article examines the judgment and its implications for the future of litigation and AI-assisted legal practice in India.
II. Factual Background and Procedural History
The dispute arose from insolvency proceedings involving Essel Infraprojects Ltd. ("EIL"), which had furnished a corporate guarantee in relation to loan facilities extended by Jammu & Kashmir Bank Ltd. to Pan India Utilities Distribution Company Ltd. ("PIUDCL"). Following the borrower's default, the bank initiated proceedings under Section 7 of the Insolvency and Bankruptcy Code, 2016 ("IBC"), before the NCLT, Mumbai.3
By an order dated 28 August 2024, the NCLT admitted the application and initiated the corporate insolvency resolution process against EIL. The appellant, Pooja Ramesh Singh, challenged the order before the NCLAT, which dismissed the appeal on 11 September 2025.4
Before the Supreme Court, the appellant challenged, inter alia, the authenticity of six judicial authorities relied upon in the NCLT's decision and subsequently referred to by the NCLAT. The authorities were alleged to be fictitious or incorrectly attributed to genuine judgments.
Upon independent examination, the Supreme Court found that the citations and extracts suffered from serious authenticity defects. Some authorities did not exist, while others involved genuine citations accompanied by non-existent paragraphs. One citation was attributable to an entirely different judgment.5
The Court further recorded that the respondent bank had filed an affidavit indicating that its counsel had not cited the disputed authorities before the NCLT. The material appeared to have been obtained through the adjudicating authority's own research. The fabricated authorities subsequently escaped scrutiny at the appellate stage.6
The case therefore raised an important question extending beyond the conduct of any individual advocate: whether a judicial determination can be sustained when its reasoning is founded upon fictitious material presented and treated as judicial precedent.
III. Issues Before the Supreme Court
The principal issues arising for consideration were:
- Whether decisions of courts or tribunals could be sustained where the adjudication relied upon fictitious or AI-generated judicial authorities.
- Whether the introduction of such material into the decision-making process compromised the integrity and legality of the resulting adjudication.
- What safeguards and professional standards were required to address the unverified use of AI-generated legal material by members of the Bar and the Bench.
The Court approached these questions by examining the authenticity of the cited authorities, the consequences of relying upon them and the need for effective control over AI-assisted adjudication.
IV. The Supreme Court's Findings and Ratio Decidendi
A. Authenticity of judicial precedent is indispensable
The Supreme Court independently examined the six authorities relied upon by the NCLT and identified substantial discrepancies. These included fictitious citations, incorrect attribution of genuine citations and purported judicial paragraphs that could not be found in the actual judgments.7
The significance of this finding lies in the distinction between an error in legal interpretation and the use of material that does not exist as judicial precedent.
An advocate may advance an argument that a court ultimately rejects. A court may also interpret a genuine judgment differently from a party. Such disagreements form part of ordinary adjudication. However, a proposition purportedly supported by a judgment that does not exist presents an entirely different problem: the adjudicator is being invited to reason from a false legal premise.
The judgment accordingly reinforces the proposition that legal authorities must be capable of independent verification. The apparent plausibility of a citation, the sophistication of the language used or the confidence with which a proposition is presented cannot substitute for authenticity.
B. A decision founded upon fabricated material cannot be sustained
The Court held that the judicial process and the decisions under challenge had been tainted by reliance upon material represented as precedent but found to be unreal or non-existent. It concluded that a decision founded upon fake or hallucinated material could not be sustained in law.8
This conclusion is central to the ratio decidendi.
The concern was not restricted to whether a particular fabricated authority independently determined the final outcome. Rather, the Court focused on the integrity of the decision-making process itself. Judicial reasoning must proceed from authentic material, and the introduction of fictitious precedent compromises the legitimacy of that process.
The Court therefore set aside the orders of the NCLT and NCLAT and restored the bank's Section 7 application to its original number for fresh consideration. It expressly clarified that it had not expressed any opinion on the merits of the insolvency dispute.9
The decision must be understood in its factual context. It addresses adjudication materially tainted by fake or hallucinated legal material; it should not be read as holding that every ordinary citation error, inadvertent typographical mistake or immaterial inaccuracy automatically invalidates a judicial order.
C. AI must remain subject to meaningful human control
The Court recognised that AI possesses substantial capabilities to assist human effort in routine and intellectual work. At the same time, it expressed concern about the possibility of professionals becoming dependent upon technological outputs to the point of delegating their own thinking and reasoning.10
The judgment distinguishes between the legitimate use of technology and the uncritical acceptance of machine-generated conclusions.
The Court did not reject technological assistance as such. Instead, it emphasised the need for deliberate choices concerning when and how AI should be used, particularly where its output enters the adjudicatory process. It observed that the effective regulation of AI requires public policy and enforceable rules, rather than judicial declarations alone.11
The underlying principle is that the use of AI does not eliminate the need for human judgment. A legal proposition generated by an AI system must still be evaluated against authentic legal sources, applicable statutory provisions and binding judicial precedents.
D. Professional accountability of advocates and judicial responsibility
The Court adopted a stringent position against citing or relying upon AI-generated fictitious authorities without verification. It characterised an advocate's unverified citation of such material as professional misconduct and a judge's reliance upon it as a serious lapse.12
These observations have direct implications for legal practice.
An advocate is responsible for the submissions made on behalf of a client, including the authenticity and accuracy of the authorities relied upon. The fact that a citation was generated by an AI tool does not, by itself, excuse its inclusion in a pleading, written submission or compilation of authorities.
Similarly, judicial independence and responsibility require adjudicators to exercise their own judgment and ensure that the material forming the basis of a decision is legally authentic.
The Court also directed the Bar Council of India to constitute a committee to consider the issue of advocates submitting fake or hallucinated material as precedent and to prescribe appropriate guiding principles and disciplinary consequences.13
The precise application of professional disciplinary measures will depend upon the applicable framework and the circumstances of the individual case. Nevertheless, the judgment makes clear that unverified reliance upon fabricated judicial authorities is a serious professional concern.
V. Implications for Legal Research and Litigation Practice
The judgment has significant practical implications for advocates, law firms, corporate legal departments and legal researchers.
A. Verification of case citations
Every judicial authority identified through AI-assisted research should be independently located in a reliable legal database or an official court repository.
The verification process should establish the correct case title, citation, court, date of decision and relevant paragraph numbers. Merely confirming that a citation resembles a recognised reporting format is insufficient.
B. Verification of judicial extracts
The existence of a judgment does not establish that every passage attributed to it is authentic.
A lawyer must examine the relevant portion of the full judgment and confirm that the extracted passage appears in the source. The passage must also be read in context to determine whether it supports the proposition for which it is cited.
This distinction is especially important because the Supreme Court identified cases in which genuine citations were associated with paragraphs that did not exist.14
C. Verification of precedential status
Even an authentic judgment may not represent the current legal position. Subsequent decisions may have distinguished, overruled or modified its reasoning, or legislation may have altered the applicable law.
Legal research should therefore examine not only the existence of an authority but also its present legal status, precedential value and applicability to the dispute.
D. Responsibility for AI-assisted pleadings
The advocate signing, settling or presenting a pleading should exercise independent professional judgment over its contents. AI-assisted drafting must not be treated as a substitute for reviewing the factual assertions, statutory references and judicial authorities incorporated into the final document.
A prudent practice is to maintain an internal record of the primary sources consulted and the verification undertaken before filing.
E. Institutional safeguards within law firms
Law firms can reduce the risk of fabricated authorities by adopting a documented verification protocol. Such a protocol may require:
- Identification of potentially relevant authorities through preliminary research.
- Retrieval of the complete judgment from an authoritative source.
- Verification of the citation, relevant paragraphs and legal proposition.
- Examination of subsequent judicial treatment and current precedential status.
- Final review of the compilation of authorities before filing or oral submissions.
These measures can improve accuracy without preventing the responsible use of AI for preliminary research, summarisation or drafting assistance.
VI. Does the Judgment Prohibit the Use of Artificial Intelligence in Legal Practice?
The judgment does not establish a blanket prohibition on the use of AI in legal research or drafting. Its concern is the use of fake or hallucinated material as though it were authentic judicial precedent.
AI tools can assist in identifying issues, generating preliminary research leads, organising documents and summarising judgments. These functions can be valuable when used with appropriate safeguards.
However, an AI-generated answer should not be treated as an authoritative source of law. The relevant proposition must be verified against the actual judgment or statutory text before it is relied upon in court.
The distinction is therefore between technological assistance and the abdication of professional responsibility. AI may facilitate the research process, but the responsibility for the accuracy of a legal submission remains with the professional who adopts and presents it.
The judgment supports a model of responsible adoption in which technological efficiency operates alongside independent verification, professional accountability and meaningful human supervision.
VII. Broader Significance for the Administration of Justice
The decision is significant beyond the immediate insolvency dispute.
Courts and tribunals necessarily rely upon the accuracy of legal materials presented during adjudication. If fictitious authorities enter the decision-making process, the resulting errors may remain undetected until an appeal or subsequent challenge. This creates avoidable litigation costs, delays and uncertainty.
The facts of the case are particularly instructive because the disputed authorities were not identified at the initial adjudicatory stage or corrected by the appellate tribunal.15 The case demonstrates that verification is a shared institutional concern, even though the professional responsibilities of individual participants remain distinct.
The Court's approach also reflects the need to preserve the integrity of legal reasoning as technology develops. Legal decision-making involves the evaluation of facts, interpretation of law and the exercise of judgment. AI-generated content can assist these functions, but cannot dispense with the requirement that the materials relied upon be genuine and legally relevant.
The direction to the Bar Council of India to constitute a committee is an important institutional response. Its effectiveness will depend upon the development and implementation of clear professional standards, appropriate verification practices and proportionate accountability mechanisms.
VIII. Conclusion
The Supreme Court's decision in Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. provides an important judicial response to the risks associated with AI-generated fictitious legal authorities.
The judgment reinforces three essential principles: judicial precedent must be authentic, adjudication must remain grounded in verifiable legal material, and the adoption of AI does not displace professional responsibility.
For advocates and law firms, the practical lesson is clear. Every authority intended for reliance in court must be independently verified, including its citation, relevant passages, legal proposition and current precedential status. The source of an error—whether manual research or an AI tool—does not remove the need for professional scrutiny.
The decision should not discourage responsible technological innovation. Rather, it demonstrates why innovation must be accompanied by safeguards that preserve accuracy, accountability and confidence in the administration of justice.
As AI becomes more deeply integrated into legal practice, the value of professional judgment will lie not merely in the ability to generate legal arguments quickly, but in the ability to distinguish reliable authority from plausible fabrication. The integrity of adjudication ultimately depends upon that distinction.
Table of Authorities
Cases
Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., Civil Appeal No. 11950 of 2025, 2026 INSC 668 (India July 2, 2026).
Statutes
Insolvency and Bankruptcy Code, No. 31 of 2016, § 7 (India).
Primary Sources
- Supreme Court of India, Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., Civil Appeal No. 11950 of 2025, 2026 INSC 668 (July 2, 2026), Full judgment (PDF).
- Supreme Court of India, Landmark Judgment Summaries, entry concerning Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., 2026 INSC 668.
Disclaimer
This article is intended for general legal information and academic discussion only. It does not constitute legal advice or create an advocate—client relationship. Readers should consult the complete judgment and applicable law before relying upon the analysis in any specific matter.
Footnotes
Footnotes
-
Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., Civil Appeal No. 11950 of 2025, 2026 INSC 668, ¶¶ 1, 14–17 (India July 2, 2026). ↩
-
Id. ¶¶ 1–9. ↩
-
Id. ¶¶ 10–13; Insolvency and Bankruptcy Code, No. 31 of 2016, § 7 (India). ↩
-
Pooja Ramesh Singh, 2026 INSC 668, ¶¶ 10–14. ↩
-
Id. ¶ 15. ↩
-
Id. ¶ 16. ↩
-
Id. ¶ 15. ↩
-
Id. ¶ 17. ↩
-
Id. ¶¶ 18–21. ↩
-
Id. ¶¶ 2–4. ↩
-
Id. ¶ 5. ↩
-
Id. ¶ 7. ↩
-
Id. ¶¶ 8–9. ↩
-
Id. ¶ 15. ↩
-
Id. ¶ 16. ↩



