Supreme Court voids NCLT insolvency orders over fake AI precedents and demands zero tolerance

The Supreme Court of India has set aside insolvency orders issued against Essel Infraprojects Limited after discovering that the adjudicatory process relied on non-existent and falsely attributed judicial precedents associated with artificial intelligence-generated legal research.

In its July 2, 2026 ruling, the Supreme Court held that producing, citing or relying on unverified artificial intelligence-generated precedents threatens the integrity of the justice system. The court said lawyers commit professional misconduct when they cite fabricated judgments without verification, while judges and tribunal members commit a serious institutional lapse if they rely on such material while deciding cases.

A bench comprising Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe set aside orders of both the National Company Law Tribunal and the National Company Law Appellate Tribunal in the Essel Infraprojects insolvency dispute. The matter has been returned to the National Company Law Tribunal for a fresh decision based on the facts and genuine legal authorities.

The Supreme Court also directed the Bar Council of India to constitute a committee to examine artificial intelligence use in legal practice, formulate guiding principles and consider disciplinary measures for advocates who place fabricated or hallucinated precedents before courts.

Why did the Supreme Court invalidate the NCLT and NCLAT orders involving Essel Infraprojects?

The appeal arose from insolvency proceedings initiated by Jammu and Kashmir Bank Limited against Essel Infraprojects Limited under Section 7 of the Insolvency and Bankruptcy Code. Essel Infraprojects had provided a corporate guarantee for credit facilities extended to Pan India Utilities Distribution Company Limited.

The National Company Law Tribunal’s Mumbai bench admitted the insolvency application on August 28, 2024, after recording an alleged default of approximately ₹87.43 crore. The National Company Law Appellate Tribunal upheld that decision on September 11, 2025, prompting suspended director Pooja Ramesh Singh to approach the Supreme Court.

During the Supreme Court proceedings, the appellant argued that the tribunal orders relied on six decisions that either did not exist or did not contain the legal propositions attributed to them. Some references used apparently genuine case names with invented citations, while others placed fabricated passages inside authentic Supreme Court decisions.

The disputed authorities reportedly included purported decisions titled State Bank of India v. Shree Ram Urban Infrastructure, Everest Kento Cylinders v. Union of India and ICICI Bank v. Urban Infrastructure Real Estate. Verification through recognised legal databases failed to establish that the cited decisions and extracts existed in the form presented before the tribunals.

The Supreme Court concluded that a judicial decision cannot be preserved after fabricated precedents enter the reasoning process. The problem was not treated as a minor citation error that could be corrected while leaving the final conclusion untouched. Instead, the court held that reliance on fake legal authority contaminated the adjudication itself.

The National Company Law Tribunal must now reconsider the insolvency dispute using the factual record and verifiable law. The Supreme Court did not finally determine whether insolvency proceedings against Essel Infraprojects should succeed or fail.

Why did the Supreme Court describe fake AI precedents as a catastrophic threat to justice?

Generative artificial intelligence systems produce text by predicting plausible responses from patterns in training data and supplied material. A response may sound legally authoritative while including a case, citation, statutory provision or judicial quotation that does not exist.

This behaviour is commonly described as an artificial intelligence hallucination. In ordinary conversation, an invented detail may be corrected without serious consequences. In litigation, a fabricated precedent can affect commercial rights, personal liberty, property, employment, taxation or the continuation of a company’s business.

The Supreme Court warned that artificial intelligence-generated falsehoods can be especially dangerous because they may appear polished and credible. A fictional case citation can resemble a genuine legal reference closely enough to escape immediate detection, particularly when judges and lawyers are handling large records under heavy time pressure.

The court compared the introduction of fabricated artificial intelligence material into adjudication with releasing a dangerous contaminant into the legal system. It described the damage as potentially invisible and insidious before its consequences become apparent.

A false precedent can spread beyond the original proceeding. Once included in an official tribunal or court order, other lawyers may repeat it, legal databases may index it and later courts may assume that an earlier authority has already verified it. One fabricated citation can therefore create a chain of apparently legitimate but fundamentally false legal reasoning.

The Supreme Court’s response was intended to stop that process at its earliest stage. The ruling establishes that apparent convenience cannot excuse the use of unverifiable authorities and that an adjudicatory institution must protect the reliability of its precedential record.

What responsibilities has the Supreme Court placed on lawyers using artificial intelligence?

The judgment makes clear that an advocate cannot defend a fabricated citation by blaming the software that produced it. Lawyers remain professionally responsible for every authority, quotation and proposition placed before a court in their name.

Artificial intelligence may assist with identifying issues, organising documents, summarising material or suggesting search terms. It does not replace the advocate’s obligation to locate the original judgment, read it and confirm that it supports the legal argument being advanced.

The Supreme Court held that citing an artificial intelligence-generated judgment without verification can amount to professional misconduct. This creates consequences beyond losing an individual case because disciplinary proceedings may affect an advocate’s professional standing and ability to practise.

Verification requires more than checking whether a case title appears somewhere online. A lawyer must confirm that the citation is accurate, that the judgment was delivered by the stated court, that it has not been overruled and that the quoted passage actually appears in the decision.

The context of a precedent is equally important. Artificial intelligence may extract a sentence while omitting the facts, procedural position or later reasoning that limits its application. A genuine quotation can therefore become misleading when separated from the judgment’s actual holding.

The ruling places responsibility on senior advocates, arguing counsel, instructing lawyers and legal teams according to their involvement. Firms and chambers using artificial intelligence tools are likely to need internal verification procedures rather than leaving each researcher to decide independently how generated material should be checked.

Why did the Supreme Court also hold judges and tribunals responsible for AI-generated errors?

The Supreme Court did not treat the issue solely as professional wrongdoing by lawyers. It said judges and adjudicators also have a duty to verify the authorities on which they rely, particularly when a citation materially affects the outcome.

Judicial reasoning cannot be outsourced to a software system or accepted from counsel without scrutiny. Courts must maintain human control over adjudication at every stage, even when technology is used to accelerate research, document review or administrative work.

This does not require judges to independently reconstruct every aspect of every party’s research. It does require greater caution when a precedent is unfamiliar, difficult to locate or presented in a form inconsistent with recognised legal databases.

The Supreme Court’s intervention is particularly important for tribunals dealing with specialised and high-volume disputes. The National Company Law Tribunal handles insolvency and corporate matters that can involve large financial claims, creditor recoveries, changes in company control and the possible liquidation of businesses.

An insolvency admission order is not merely procedural. It can suspend the authority of a company’s existing board, place the business under an insolvency professional and trigger a statutory process involving creditors, employees and investors.

Using a fictional precedent in such a decision creates a direct risk of economic harm. The Supreme Court therefore concluded that even a small amount of fabricated authority entering judicial reasoning can make the determination legally unsustainable.

The ruling may encourage courts to insist on copies of cited decisions, verified database references and clearer identification of the paragraphs supporting each proposition. Digital court systems may also require tools that flag invalid citations without allowing automated checking to become a substitute for judicial responsibility.

What has the Bar Council of India been directed to do after the Supreme Court ruling?

The Supreme Court directed the Bar Council of India to create a committee to examine the use of artificial intelligence-generated material by lawyers and develop appropriate guiding principles.

The committee is expected to consider how artificial intelligence may be used legitimately while preventing fabricated authorities from reaching courts. The process could lead to professional standards covering disclosure, citation verification, supervision of junior researchers and accountability for documents produced with automated tools.

The Bar Council of India may also consider disciplinary consequences when lawyers submit false authorities. The Supreme Court’s finding that such conduct can constitute misconduct gives the regulatory body a clear basis for developing enforceable rules.

Any guidelines will need to distinguish between several forms of artificial intelligence use. Using software for grammar correction does not create the same risk as asking a generative system to identify precedents and copying its response into written submissions.

Similarly, a legal research tool connected to a verified database may carry lower hallucination risks than a general-purpose chatbot, although no automated system removes the need for professional review.

The Bar Council of India must also consider whether lawyers should disclose artificial intelligence assistance. Mandatory disclosure for every minor use could become impractical, while no disclosure requirement may make it difficult to investigate how fabricated authorities entered a case.

Training may be as important as punishment. Many legal professionals use new tools without understanding how confidently they can produce false information. Clear verification protocols could prevent misconduct before disciplinary proceedings become necessary.

How could the judgment reshape artificial intelligence use across India’s legal system?

The immediate result is limited to the Essel Infraprojects dispute, but the principles announced by the Supreme Court apply much more broadly.

Indian courts and tribunals are experimenting with artificial intelligence for translation, transcription, legal research, case management and document analysis. These applications could reduce delays and improve access to information in a judicial system handling an enormous volume of litigation.

The Supreme Court has not rejected that technological transition. Instead, the ruling draws a boundary between artificial intelligence used as an aid and artificial intelligence treated as an authority.

A system may help locate potentially relevant judgments, but the verified judgment remains the source of law. A tool may summarise a document, but the court must decide the case from the original record. A generated draft may save time, but the judge or lawyer remains responsible for every statement it contains.

The ruling could encourage legal technology companies to build stronger safeguards. Products designed for Indian lawyers may need citation links, source-document access, warning labels, database verification and audit records showing how an answer was generated.

Law firms may respond by creating artificial intelligence policies that require researchers to preserve source documents and independently validate every cited proposition. Courts may introduce filing requirements under which advocates certify that precedents in their submissions have been verified.

Law schools and judicial academies may also need to teach artificial intelligence literacy. Knowing how to prompt a system is less important than recognising when an answer is unreliable and understanding the ethical obligations that remain with the human user.

Why does the ruling matter beyond one insolvency case and one artificial intelligence error?

The credibility of a legal system depends on parties believing that decisions are based on real statutes, genuine precedents and evidence available for examination.

A fabricated authority prevents meaningful challenge because the opposing party cannot read, distinguish or appeal against a judgment that never existed. It also damages the principle of open justice because the supposed basis for the decision cannot be independently inspected.

The Essel Infraprojects case demonstrates that artificial intelligence errors are no longer theoretical risks confined to technology demonstrations. Fabricated material entered formal adjudication, appeared in tribunal reasoning and survived an appellate stage before reaching the Supreme Court.

That progression explains the severity of the ruling. The Supreme Court was not addressing a lawyer’s private research note but an error that had become embedded in official orders affecting a company facing insolvency proceedings.

The judgment also arrives as businesses, governments and courts are under pressure to adopt artificial intelligence quickly. Institutions may believe that automated tools can reduce workloads, but speed without verification can create larger delays when defective orders are later challenged and set aside.

The broader lesson is that artificial intelligence governance in the justice system cannot focus only on privacy, cybersecurity and procurement. It must also protect the authenticity of legal authorities and the human accountability behind every decision.

What happens next in the Essel Infraprojects case and India’s legal AI debate?

The insolvency application against Essel Infraprojects will return to the National Company Law Tribunal for fresh adjudication. The tribunal must assess Jammu and Kashmir Bank’s claim, the corporate guarantee, the alleged default and the parties’ legal arguments without relying on the fabricated authorities identified during the Supreme Court proceedings.

The remand does not automatically end the insolvency case. Jammu and Kashmir Bank may still succeed if the statutory requirements under the Insolvency and Bankruptcy Code are established through admissible evidence and genuine legal precedent.

The Bar Council of India’s committee process will determine whether the ruling leads to detailed national standards. Its recommendations could influence lawyers practising before the Supreme Court, High Courts, district courts and specialised tribunals.

Further questions are likely to arise over responsibility when a junior lawyer, external vendor or artificial intelligence research platform generates a false citation. Professional rules will need to establish who must verify the result and how accountability should be distributed within legal teams.

The judiciary may also examine technological solutions, but automated citation verification must itself be treated cautiously. A reliable system can assist human reviewers, while an opaque or incomplete database can create a new layer of false confidence.

The Supreme Court has nevertheless established the core principle clearly: artificial intelligence may support legal work, but adjudication must remain under complete human control.

What are the key takeaways from the Supreme Court ruling on fake AI legal precedents?

• The Supreme Court of India set aside National Company Law Tribunal and National Company Law Appellate Tribunal orders in the Essel Infraprojects insolvency case after fabricated and non-existent authorities entered the adjudicatory reasoning.
• The July 2, 2026 judgment was delivered by Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe in an appeal filed by suspended Essel Infraprojects director Pooja Ramesh Singh.
• The insolvency proceedings originated from a Section 7 application filed by Jammu and Kashmir Bank Limited involving an alleged default of approximately ₹87.43 crore and a corporate guarantee issued by Essel Infraprojects.
• The Supreme Court held that an advocate who cites an artificial intelligence-generated judgment without verifying its existence and legal relevance may commit professional misconduct.
• Judges and tribunal members also carry responsibility because relying on fake or hallucinated authority is a serious adjudicatory lapse that can invalidate the resulting decision.
• The Essel Infraprojects matter has been remanded to the National Company Law Tribunal for fresh consideration based on the evidence and genuine, verifiable legal precedents.
• The Bar Council of India has been directed to constitute a committee, formulate guiding principles for legal artificial intelligence use and consider disciplinary action involving fabricated authorities.
• The ruling permits artificial intelligence to assist legal research and administration but insists that lawyers and judges retain complete human control and responsibility throughout the adjudicatory process.