Supreme Court’s Zero-Tolerance Rule on AI-Hallucinated Case Law: Lessons for NCLT & NCLAT Practice

Artificial intelligence is increasingly used in legal research, insolvency strategy and professional drafting. The Supreme Court of India has now drawn a sharp line around that use. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., 2026 INSC 668, decided on 2 July 2026, the Court set aside orders of the NCLT and NCLAT after fake and AI-hallucinated precedents entered the decision-making process. For professionals appearing before the NCLT and NCLAT, the judgment is not merely about technology risk. It is about the integrity of adjudication, verification standards and professional accountability.

Background: how an ordinary Section 7 insolvency case became an AI-governance judgment

The appellant was a suspended director of Essel Infraprojects Ltd. (EIL). EIL had given a corporate guarantee for facilities extended by Jammu and Kashmir Bank Ltd. to Pan India Utilities Distribution Company Ltd. After default, the bank invoked the insolvency framework and filed an application under Section 7 of the Insolvency and Bankruptcy Code, 2016 against EIL as corporate debtor and guarantor. Section 7 permits a financial creditor to initiate the corporate insolvency resolution process when a default has occurred. The NCLT, Mumbai admitted the application on 28 August 2024, appointed an Interim Resolution Professional and declared the statutory moratorium.

The admission order was challenged before the NCLAT. On 11 September 2025, the NCLAT dismissed the appeal. The controversy reached the Supreme Court when it was demonstrated that six authorities reproduced in the NCLT/NCLAT reasoning were problematic: some citations were entirely non-existent, while others were genuine citations attributed with paragraphs that could not be found in the actual judgments.

What the Supreme Court found

The Supreme Court independently examined the authorities. Of the six citations, two were real cases but were paired with non-existent paragraphs; one citation belonged to a different real judgment but was again attributed with a non-existent paragraph; and the remaining three citations did not exist at all. Significantly, the respondent bank filed an affidavit stating that these judgments had not been cited by its counsel at the Bar. The material therefore appeared to have entered the order through the adjudicatory research process itself, and the error subsequently escaped scrutiny before the NCLAT.

The core rule: verification is non-negotiable

The Court adopted what it described as a zero-tolerance approach to fake or hallucinated precedent. It held that an advocate who presents such material without verification commits misconduct, while judicial reliance on it is an equally serious lapse. The Court went further: where fake or hallucinated material enters the decision-making process, the resulting decision is no decision in law, irrespective of whether the false material ultimately changed the outcome. The defect goes to the sanctity and legitimacy of adjudication itself.

At the same time, the Court did not prohibit legitimate use of AI. The judgment recognizes that technology can materially improve efficiency. The legal risk arises when AI output is treated as authority without independent human verification. In professional practice, AI may assist research; it cannot replace source checking.

Result in the insolvency proceedings

Because the NCLT and NCLAT decisions had been tainted by non-existent authorities, the Supreme Court set them aside. The bank's Section 7 application was restored to its original number and remitted to the NCLT for fresh consideration on merits. The Supreme Court therefore did not allow an insolvency admission order to survive merely because the underlying debt or default issues might independently have supported it; the adjudicatory process itself had to be legally clean.

Direction to the Bar Council of India

The Court also directed the Bar Council of India to constitute a committee to consider the problem of advocates submitting fake or hallucinated authorities and to prescribe guiding principles and disciplinary consequences. This is important for law firms, insolvency teams and professional services organisations because AI-use controls may increasingly become part of professional-risk frameworks rather than remaining an informal internal practice.

Practical implications for NCLT and NCLAT matters

First, every case citation used in a petition, reply, written submission, legal opinion or internal note should be checked against an authoritative database or the court's own judgment record. A citation alone is not enough; the proposition and paragraph relied upon must also be verified.

Second, insolvency professionals and financial creditors should apply the same discipline to legal memoranda generated for Section 7, Section 9, avoidance-transaction, guarantor, resolution-plan and liquidation disputes. A hallucinated precedent can contaminate not only courtroom submissions but also committee papers, lender approvals and litigation strategy.

Third, firms should create an AI research protocol. The person who uses an AI tool should identify the underlying source, confirm the neutral citation or reported citation, open the original judgment, verify the exact proposition and retain a source copy in the matter file. High-risk propositions should receive a second-level review.

Fourth, professionals should distinguish between AI-assisted summarisation and legal authority. A generated summary can be a useful starting point, but only the statute, rule, regulation, circular, notification or judgment itself should be treated as the operative legal source.

A governance checklist for professional firms

For CA firms, law firms, insolvency professionals, lenders and corporate legal departments, the judgment supports a simple control framework: require source-level verification of all authorities; prohibit filing AI-generated citations without validation; record who verified each important precedent; maintain copies of primary sources; conduct partner or senior review for contentious legal propositions; and train teams to recognise that plausible language is not evidence of legal accuracy.

Why this ruling matters beyond insolvency law

The judgment arose from an IBC proceeding, but its reasoning is broader. Tax litigation, company-law petitions, regulatory appeals and writ proceedings all depend on reliable precedent. As AI tools become more capable, professional advantage will not come merely from faster drafting. It will come from combining speed with disciplined verification. In regulated professional work, provenance of information is becoming as important as the information itself.

Key takeaway

Pooja Ramesh Singh is now a critical professional-risk precedent. AI can assist legal and insolvency work, but responsibility cannot be delegated to the tool. Before a case reaches a tribunal or court, every authority must be real, every quoted proposition must be traceable and every legal conclusion must remain under human professional control.

Primary sources

Supreme Court of India: Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., Civil Appeal No. 11950 of 2025, 2026 INSC 668, judgment dated 2 July 2026. Statutory framework: Section 7 of the Insolvency and Bankruptcy Code, 2016 as published on India Code. The Supreme Court's official landmark judgment summary records the factual background, verification findings, zero-tolerance principle, remand to the NCLT and direction to the Bar Council of India.

Disclaimer

This article is intended for professional information and general discussion only. It is not legal, tax or insolvency advice. Facts and procedural strategy should be evaluated against the complete judgment, applicable law and the circumstances of each matter before taking any action.