Published On: October 14, 2026
Authored By: Angel Singh
IILM University
Abstract
The intersection of legal technology and judicial integrity requires an urgent re-examination of lawyerly ethics and evidentiary standards in India. Legal research once involved searching databases, using keywords, reading numerous judgments, identifying relevant authorities, checking whether a precedent is still good law, and extracting the legal principle. The recent ruling in Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (2026)[1] serves as a critical case study on the dangers of integrating generative AI into legal research. This article analyses how AI-hallucinated case law challenges the traditional doctrine of stare decisis within the Indian judicial system. What happens to the administration of justice when the precedent cited before a judge exists only in the digital imagination of an AI model? Can the Indian judiciary protect the integrity of its decisions when artificial intelligence begins to invent its own rules of law?
The research objective of this article is to examine whether reliance on AI-generated, non-existent judicial precedents undermines the integrity of judicial decision-making, and what safeguards are necessary to prevent such occurrences in India. The core problem is what happens when a lawyer or an adjudicatory body relies upon a case that AI has invented. AI helps lawyers prepare the first version or structure of legal documents such as legal notices, written submissions, petitions and affidavits, and it can convert notes into a structured format while identifying missing arguments. The problem, however, is not the use of AI itself, but the uncritical acceptance of AI-generated legal information.
Keywords: legal research; AI-hallucinated case law; judicial integrity; stare decisis; Indian judicial system.
I. Introduction
The integrity of Indian jurisprudence faces an unprecedented threat, not from human error, but from the flawed algorithms of artificial intelligence. When a computer algorithm fabricates legal history, the foundational pillars of judicial precedent begin to crumble. The boundary between technological innovation and legal fiction has officially blurred in the Indian courtroom. This article analyses how AI-hallucinated case law challenges the traditional doctrine of stare decisis within the Indian judicial system. As Indian courts increasingly rely on digital tools, Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (2026) exposes the severe systemic risks of AI-generated legal fabrications. This article also explores the urgent regulatory and ethical frameworks needed to combat the rise of fake AI-generated precedents in Indian litigation.
Traditionally, legal research involved searching databases, using keywords, reading numerous judgments, identifying relevant authorities, checking whether a precedent is still good law, and extracting the legal principle. With the emergence of a new era, however, everything is shifting from the traditional to the digital. AI changes this process by allowing lawyers to use natural-language questions instead of relying entirely on keywords. Nowadays, people have grown overly dependent on AI even for basic things, which is making us digitally dependent, and that is not desirable.
II. What Is an AI Hallucination?
An AI hallucination occurs when a generative AI model produces content that appears plausible and authoritative but is inaccurate or entirely fabricated. AI has transformed legal research from a predominantly keyword-driven exercise into an increasingly semantic and conversational process. AI-powered legal research platforms can identify conceptually similar precedents, summarise judgments, trace citation relationships, and assist lawyers in locating authorities relevant to a particular legal proposition. AI can similarly assist in drafting by organising facts, generating preliminary structures and refining legal language.
However, these advantages do not eliminate the lawyer’s duty to verify every authority independently. In legal practice, an apparently precise answer is not necessarily an accurate one. The ability of generative AI to produce plausible but fabricated authorities creates a fundamental tension: the technology that makes legal research faster can, when unchecked, make legal misinformation faster as well. The problem is not the use of AI itself, but the uncritical acceptance of AI-generated legal information.
III. The Major Development: Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd.
Jammu and Kashmir Bank Ltd. extended credit of ₹200 crores to the principal borrower, Pan India Utilities Distribution Company Ltd. (PIUDCL).[2] The facility was secured by a corporate guarantee from Essel Infraprojects Ltd. (EIL), which created a mortgage over 196.16 acres of land in Gorai village, Borivali (West), Mumbai, owned by EIL itself.[3] On 18 November 2017, a renewal-cum-reduction letter was issued, reducing the facility limit to ₹130 crores and stipulating that all other existing terms and conditions would remain applicable.
Following continuous defaults by the principal borrower, the Bank initiated the Corporate Insolvency Resolution Process (CIRP) against EIL, the corporate debtor, by filing an application under Section 7 of the Insolvency and Bankruptcy Code, 2016. On 28 August 2024, the National Company Law Tribunal (NCLT), Mumbai, found the existence of both debt and default, admitted the application, appointed an Interim Resolution Professional (IRP) and declared a statutory moratorium.
Aggrieved by the admission order, EIL’s suspended director, Pooja Ramesh Singh, appealed to the National Company Law Appellate Tribunal (NCLAT). She contended that the restructuring of EIL, especially its demerger and subsequent amalgamation, had legally transferred EIL’s liabilities, thereby extinguishing its guarantee.
On 11 September 2025, the NCLAT dismissed the appeal and upheld the NCLT’s order. It affirmed the NCLT’s reasoning, which relied on six Supreme Court precedents that were purported to exist.[4] When the dispute reached the Supreme Court, this procedural mistake surfaced. Senior Counsel for the Appellant pointed out that the six citations were fake, non-existent and completely AI-hallucinated.[5] Subsequently, counsel for the financial creditor, Jammu and Kashmir Bank Ltd., submitted an affidavit clarifying that the Bank had never cited these fabricated cases and that the research had instead been conducted by the adjudicating authority at the NCLT through unverified AI.
IV. Legal Issues Raised
1. The Rule of Law
When a tribunal justifies deprivation, state coercion or the initiation of corporate insolvency using fake, non-existent precedents, it commits an act of judicial arbitrariness. When adjudicating authorities rely on such “phantom laws”, never passed by a legislature nor tested in constitutional courts, they threaten the very foundation of justice in its true sense. Therefore, the Supreme Court’s insistence on setting aside the tribunal’s decision, without first asking whether the fake cases actually changed the final outcome, is not a new innovation.
2. Natural Justice
Natural justice, specifically the maxim audi alteram partem (let the other side be heard), refers to the principles of fairness, equity and justice. It grants litigants access to a fair hearing and acts as a shield against arbitrary, biased and unfair administrative practices. Fabricated cases destroy the adversarial process. A lawyer can argue that a real precedent was wrongly decided or is inapplicable, but phantom cases deprive litigants of the opportunity to read, interpret and rebut them during hearings.
3. Judicial Accountability
Under the separation of powers, the power of judicial review and the dispensation of justice lie exclusively with the human officers of the judiciary. Articles 124(3)[6] and 217(2)[7] of the Constitution provide that a person shall not be qualified for appointment as a judge unless they are a “citizen of India”, and citizenship is a legal status granted exclusively to natural persons under the Citizenship Act, 1955.[8] An algorithm, AI model or software program cannot be a citizen of India, nor can it hold a constitutional office. It cannot sign a judgment, take an oath, or be impeached by Parliament under Article 124(4). Hence, the Constitution demands identifiable human decision-making, because accountability intrinsically requires moral agency.
The Supreme Court, in adjudicating this matter, took the profound philosophical stance of “Saadhana”,[9] the struggle to arrive at the truth. It asserts that the intellectual struggle of judging is a non-delegable duty.
V. Conclusion
Legal research has moved from keyword searches and manual reading to natural-language queries and AI-generated answers. This shift brings speed, but it also brings a new risk: the fabrication of legal history. Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. shows how this risk can enter a tribunal’s reasoning and be carried forward, unchecked, by an appellate forum.
When a tribunal relies on non-existent precedents, the Rule of Law is weakened by arbitrariness, natural justice is denied because litigants cannot rebut cases that do not exist, and judicial accountability is lost because only a human being can be answerable for a judgment. The foundational pillars of judicial precedent, and with them the doctrine of stare decisis, depend on authorities that are real, verifiable and human-vetted.
Lastly, I would like to say: “AI may assist in finding the law, but it cannot be permitted to invent the law.”
References
[1] Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., 2026 INSC 668 (India).
[2] Id.
[3] Id.
[4] Order of the National Company Law Appellate Tribunal dated 11 Sept. 2025, ¶ 12, discussed in Pooja Ramesh Singh, 2026 INSC 668.
[5] Pooja Ramesh Singh, 2026 INSC 668.
[6] INDIA CONST. art. 124, cl. 3.
[7] INDIA CONST. art. 217, cl. 2.
[8] Citizenship Act, No. 57 of 1955, INDIA CODE (1955).
[9] Pooja Ramesh Singh, 2026 INSC 668.
