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

Citation: 2026 INSC 668 | Civil Appeal No. 11950 of 2025 | 2026 (SC) 653

Decided: 2 July 2026

Bench: Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe

Introduction

What happens when a court relies on judgments that do not exist?

On 2 July 2026, the Supreme Court of India faced this question in the starkest possible form. It found that the National Company Law Tribunal (NCLT) had used fake, non-existent judicial precedents, apparently generated by an Artificial Intelligence (AI) tool, as the basis for its decision. The National Company Law Appellate Tribunal (NCLAT) then confirmed this order without noticing the fraud.

The Supreme Court's response was immediate and unequivocal. It set aside both orders. It declared that a decision based even partially on fake or hallucinated AI-generated material is no decision in the eyes of the law. It declared a policy of zero tolerance. It directed the Bar Council of India to constitute a committee to address this problem. And it used one of the most powerful comparisons in recent judicial writing: relying on AI-hallucinated precedents in judicial proceedings is like releasing methyl isocyanate, the deadly gas from the Bhopal disaster, into the province of law and justice.

Background: The Insolvency Case Behind the Controversy

The case arose from a Section 7 application under the Insolvency and Bankruptcy Code, 2016. Section 7 allows a financial creditor to initiate corporate insolvency resolution proceedings against a corporate debtor.

Jammu and Kashmir Bank Limited (J&K Bank) had extended a credit facility of Rs. 200 crore to Pan India Utilities Distribution Company Limited, a group entity. The facility was secured by a corporate guarantee from Essel Infraprojects Ltd. (EIL) and a mortgage over land at Gorai, Borivali, Mumbai.

When the loan went into default, J&K Bank filed a Section 7 application before the NCLT (Mumbai Bench), claiming an outstanding debt of Rs. 87.43 crore from Essel Infraprojects as the guarantor.

Pooja Ramesh Singh, a suspended director of Essel Infraprojects, contested the application. Her core argument was:

• Through a series of corporate restructurings, including a demerger and an amalgamation approved by the Bombay High Court in 2014, the relevant assets and liabilities of EIL had been transferred to separate group entities.
• A new sanction letter dated 18 November 2017 did not expressly continue the corporate guarantee, meaning it had been released.
• EIL was therefore not liable for the debt.

The NCLT's Decision: Where the Problem Began

The NCLT Mumbai Bench, on 28 August 2024, rejected these arguments and admitted the insolvency application. It relied upon several judicial precedents in support of its reasoning.

When Pooja Ramesh Singh challenged this before the NCLAT, the NCLAT upheld the NCLT order on 11 September 2025, relying again on what appeared to be a series of Supreme Court precedents cited in paragraph 12 of its order.

Pooja Ramesh Singh then filed Civil Appeal No. 11950 of 2025 before the Supreme Court.

The Shocking Discovery Before the Supreme Court

When the Supreme Court Bench examined the case records, the appellant's counsel demonstrated something that stopped the proceedings entirely.

The judicial precedents that the NCLT had relied upon in dismissing her arguments were not real. When the citations were independently verified, the Supreme Court found that:

• Several of the cited judgments simply did not exist. The citations were completely fabricated.
• Other citations corresponded to genuine cases, but the paragraphs extracted from them and relied upon in the NCLT order did not actually appear in those cases. The genuine citations had been used as shells, with AI-generated fictional paragraphs inserted inside them.
• The CDRs produced in the court were in editable format, further raising questions about the chain of custody.

The J&K Bank filed an affidavit before the Supreme Court revealing a further disturbing detail: the fake, non-existent judgments had not been cited by J&K Bank's counsel at the bar. The NCLT had sourced these "precedents" through its own independent research, apparently using an AI tool.

This meant the fraud had originated not from a party's lawyer but from the tribunal's own research process. The NCLAT had then confirmed this order without verifying a single citation.

Enroll in the Best Judiciary Courses at Aashayein Judiciary and prepare confidently for Judiciary, APO, and ADPO exams with expert guidance. 

The Supreme Court's Analysis and Holdings

1. AI Can Assist, But Human Control Must Be Total and Absolute

The Court began by acknowledging the role and value of AI in legal adjudication. It observed that AI is a useful and powerful tool that can make legal research faster and help lawyers and judges work more efficiently.

However, the Court drew a firm line: AI cannot replace human thinking, sound judgment or judicial discretion. It observed that deciding cases requires experience, careful reasoning and an understanding of justice; qualities that only a human mind can possess.

The Court stated its fundamental position clearly: it resolved to adopt AI technology in aid of adjudication, while at the same time asserting and declaring total and absolute control over adjudication, with a human in the loop at every stage.

2. AI Hallucinations Are Catastrophically Dangerous in Legal Proceedings

The Court addressed the phenomenon of AI hallucination: the tendency of large language model AI tools to generate plausible-sounding but entirely fictitious information, including fake case citations, non-existent legal propositions, and fabricated judicial paragraphs.

In the legal context, AI hallucination is particularly dangerous because:

• Courts implicitly trust citations placed before them by advocates. Verifying every single citation independently would be an impossible burden.
• A fake citation presented with confident formality and accurate-seeming citation details may escape detection at multiple levels of the judicial process.
• Once a fabricated precedent enters the reasoning of a court order, it taints the entire decision.

The Court used the Bhopal disaster as a metaphor. The production of fake, non-existent, and hallucinated material and its utilisation as precedents in law, is like the release of methyl isocyanate in the province of law and justice.The reference is precise and powerful: methyl isocyanate, the toxic gas that leaked from the Union Carbide plant in Bhopal on the night of 2-3 December 1984, killed thousands silently. AI hallucinations in judicial proceedings can kill the rule of law with the same silent devastation.

3. A Decision Based on Fake AI Material Is No Decision at All

The Court enunciated a powerful legal principle:"A decision of a Court or an adjudicating authority based on material which is fake and hallucinated is no decision at all, and it amounts to subversion of the rule of law. Such a decision is unsustainable and has to be set aside at the earliest."

This is not merely a remedial direction in this specific case. The Court stated it as a general principle: if fake or hallucinated AI-generated material enters the decision-making process at any level, the resulting decision is legally void, irrespective of whether the fake material had a direct or indirect bearing on the outcome.

The Court went further.Such decisions are to be set aside even if an iota of fake or hallucinated material enters the decision-making process, as it would violate the sanctity of adjudication.

4. Zero Tolerance for Bar and Bench

The Court declared an absolute and express policy of zero tolerance in both directions: for advocates and for judges.

For AdvocatesIt is a misconduct on the part of an advocate to cite such judgments without verification. Equally, it is a serious lapse if a judge relies on such a fake or hallucinated AI-generated material as precedents in support of the determination.

Citing an AI-generated fake precedent before a court, without independently verifying its existence, amounts to professional misconduct. It does not matter whether the advocate was unaware that the citation was fake. The obligation to verify is unconditional.

Learn how the Supreme Court interpreted crime scene re-enactment and Article 20(3) in the landmark State of Tamil Nadu v. Ponnusamy & Others (2026) judgment. 

For Judges and Tribunals: Relying on such material, whether sourced from a party or through the court's own research, constitutes a serious lapse that vitiates the decision.

5. It Does Not Matter Whether the Fake Citation Came from a Party or the Tribunal's Own Research

One of the most important clarifications in the judgment is this: the source of the fake material does not mitigate the damage to the rule of law.

In this case, the J&K Bank's counsel had not cited the fake precedents. The NCLT itself had used AI for research and sourced them through that process. The appellant's counsel, in an unusual turn, was the one who exposed the fake citations.

The Court held that regardless of the source, the consequence is the same. A decision contaminated by fake AI precedents, whether placed before the court by a lawyer or generated by the court itself, is unsustainable.

6. Direction to the Bar Council of India

The Court directed the Bar Council of India (BCI), as the apex statutory body governing the legal profession, to:

• Constitute a committee to specifically deliberate on the issue of members of the bar submitting fake and hallucinated AI-generated material before courts.
• Take up this issue with utmost seriousness.
• Deliberate earnestly and prescribe guiding principles to prevent such occurrences.
• Prescribe disciplinary action that will follow a violation of the norms.

This is a significant institutional direction. It places the responsibility for regulating AI use in legal practice squarely on the bar's own statutory body, rather than leaving it to individual courts to deal with on a case-by-case basis.

Take the first step toward your judicial career. Contact for Online Judiciary Courses and prepare with experienced faculty, mock tests, and structured learning. 

The Final Orders

The Supreme Court set aside both the NCLT order dated 28 August 2024 and the NCLAT order dated 11 September 2025.

The Section 7 insolvency application was restored to its original number before the NCLT Mumbai Bench. The NCLT was directed to take up and dispose of the application expeditiously, preferably within two weeks, without being influenced by any of the earlier findings. The parties were directed to maintain status quo pending disposal.

The Supreme Court expressly declined to express any view on the merits of the insolvency dispute. The setting aside was entirely on the ground of integrity of process, not on the ground that J&K Bank's claim was wrong.

Why This Judgment Matters

It Is a Historic First in Indian Legal History

While courts in the United States (Mata v. Avianca, 2023) and the United Kingdom (R (Ayinde) v. London Borough of Haringey, 2025) had earlier confronted AI hallucination by lawyers, Pooja Ramesh Singh is the first Indian case where a tribunal itself, not a party's counsel, sourced AI-hallucinated material as part of its research. The Supreme Court's response is therefore the most direct judicial confrontation with AI misuse in judicial research in India.

It Creates a New Category of Legal Nullity

The principle that a decision contaminated by even an iota of AI-hallucinated material is no decision in the eyes of the law is a new principle of adjudication. It applies to all courts and tribunals, not just the NCLT. Any future case where such contamination is discovered must be set aside, regardless of whether the fake material was outcome-determinative.

It Places Human Accountability at the Centre of the AI Age

The Court's insistence on "human in the loop at every stage" is more than a technology governance principle. It is a constitutional statement: the rule of law depends on humans taking personal responsibility for every fact, every citation, and every proposition placed before a court. Delegation to AI does not transfer or dilute that responsibility.

It Will Shape Bar Council and Court Regulations on AI

The direction to the BCI to constitute a committee is the beginning of a regulatory framework for AI in Indian legal practice. This judgment will influence how the BCI, the State Bar Councils, and ultimately the courts themselves regulate the use of AI research tools by lawyers and tribunals.

It Connects to the Broader Question of AI and Justice

The case raises questions that judiciary exam aspirants should be able to discuss: what safeguards should regulate AI use in courts? Should courts use AI for their own independent research? What obligations do advocates have when using AI for legal research? The Pooja Ramesh Singh judgment is the starting point for all of these discussions in Indian law.

Conclusion

Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. is a judgment for the age we are living in. It says that technology is welcome in courts. But technology does not carry professional responsibility, personal accountability, or constitutional obligation. Those belong to humans, and they always will.

Frequently Asked Questions

Q1. What did the Supreme Court decide in Pooja Ramesh Singh v. J&K Bank (2026)?

 The Supreme Court set aside orders of the NCLT and NCLAT after discovering that the NCLT had relied on non-existent, AI-hallucinated judicial precedents in its decision. The Court declared that any decision based on fake or hallucinated AI-generated material is no decision in the eyes of the law, and ordered the NCLT to rehear the Section 7 IBC application afresh.

Q2. What is an AI hallucination in the legal context?

AI hallucination refers to the tendency of artificial intelligence language models to generate plausible-sounding but entirely fictitious content, including non-existent case citations, fabricated judicial paragraphs wrongly attributed to real cases, and invented legal propositions. In this case, the NCLT's own research produced such hallucinated precedents, which then formed the basis of its order.

Q3. What responsibility does an advocate have when citing AI-generated research?

The Supreme Court held that it is professional misconduct for an advocate to cite AI-generated judgments before a court without independently verifying their existence and accuracy. The obligation to verify is unconditional. Ignorance of the hallucination is not a defence. The Court also directed the Bar Council of India to constitute a committee to prescribe formal guidelines and disciplinary consequences.

Q4. Does the judgment ban the use of AI in legal research?

 No. The Court expressly clarified that the judgment shall have no bearing on the rightful use of AI as an aid to legal research. AI is welcome as a tool to improve efficiency. What the judgment prohibits is the presentation or reliance on AI-generated material as if it were a verified court precedent without independent verification. Human control over adjudication must remain total and absolute at every stage.

Q5. What is Section 7 of the Insolvency and Bankruptcy Code, 2016?

Section 7 IBC allows a financial creditor to file an application before the NCLT to initiate corporate insolvency resolution proceedings against a corporate debtor. A financial creditor is a person to whom a financial debt is owed, such as a bank that has extended a loan. In this case, J&K Bank filed a Section 7 application against Essel Infraprojects as a guarantor of a credit facility for a group entity.