Supreme Court Addresses Use of AI-Generated Judicial Precedents in Insolvency Proceedings

Supreme Court Establishes Zero-Tolerance Rule Against AI-Generated Judicial Precedents in Insolvency Cases

The Supreme Court has issued a direct warning to the legal profession about the use of artificial intelligence to generate fake judicial precedents, holding that courts cannot rely on unvalidated AI-created judgments in adjudication. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr. (2026 INSC 668; Civil Appeal No. 11950 of 2025), a Bench comprising Justice Pamidighantam Sri Narasimha and Justice Alok Aradhe delivered the judgment on July 2, 2026, addressing a critical gap in how the legal system handles artificial intelligence-assisted research and drafting.

The case arose from a National Company Law Tribunal order made under Section 7 of the Insolvency & Bankruptcy Code, 2016, which had been partly based on judicial precedents that did not actually exist. The NCLT had relied on these AI-generated citations as if they were authentic judgments from established courts. When the matter reached the Supreme Court on appeal, the Bench discovered the fabrication and took the opportunity to formulate binding principles governing the use of AI-generated documents in judicial proceedings.

The Supreme Court's response was unambiguous. Rather than simply setting aside the flawed NCLT order, the Bench adopted what it termed a zero-tolerance stance against the use of unvalidated AI-generated judgments in any form of adjudication. This approach reflects growing judicial concern about the reliability and authenticity of documents produced by machine learning systems when applied to legal research and precedent identification. The Court recognized that while AI tools may assist legal professionals, the output must be independently verified against actual court records before being presented as evidence of established law.

The decision carries immediate implications for insolvency proceedings, where detailed legal reasoning and citation of precedent are essential to the tribunal's decision-making process. An NCLT order that rests partly on fabricated precedents is fundamentally compromised because the tribunal may have been led to adopt a legal principle that does not actually exist in Indian jurisprudence. This undermines the integrity of the insolvency resolution process and exposes litigants to orders based on false legal foundations. The Supreme Court's intervention ensures that future NCLT decisions will be scrutinized for authenticity of cited authority.

Beyond the immediate case, the Bench directed the Bar Council of India to examine the broader issue of professional conduct in relation to the use of unvalidated AI-generated documents. This directive signals that the Supreme Court views the submission of fake precedents as not merely a factual error but as a potential breach of professional ethics. Advocates and solicitors who present AI-generated citations without verification may face disciplinary action under the Bar Council's conduct rules. The Bar Council will now be tasked with developing clear guidelines about how legal professionals may use AI tools, what verification steps are mandatory, and what sanctions apply to those who fail to validate AI output before filing or reliance.

The judgment does not ban the use of AI in legal practice. Rather, it establishes that any AI-generated document, including research summaries, case citations, or precedent lists, must be independently verified by the lawyer before it is used in court proceedings or submitted to any tribunal. The Court's language makes clear that the burden of verification rests entirely with the legal professional, not with the court or tribunal that receives the document. A lawyer cannot defend reliance on a fake precedent by claiming the AI tool made an error. The responsibility for accuracy remains with the person submitting the material.

This ruling addresses a growing practical problem in the legal profession. As AI-powered legal research tools have become more accessible and affordable, some practitioners have begun using them as shortcuts to case law research without understanding the limitations of these systems. Large language models, which form the backbone of many AI legal tools, can generate plausible-sounding case citations and judgment summaries that appear authentic but may be entirely fabricated. These "hallucinations," as they are known in AI terminology, are particularly dangerous in law because a false precedent can mislead a court into adopting incorrect legal principles.

The Supreme Court's decision in Pooja Ramesh Singh also has indirect significance for how courts will assess arguments based on AI-generated materials. If a party relies on a precedent that appears questionable, the opposing party or the court itself may now demand independent verification. This could lead to more rigorous scrutiny of citations in written submissions and oral arguments. Courts may begin asking advocates directly whether they have independently verified every precedent cited, particularly in complex commercial cases where multiple judgments are cited.

For the insolvency profession specifically, the ruling is particularly important. Insolvency practitioners, resolution professionals, and insolvency lawyers often work under tight timelines and with large volumes of case law to review. The temptation to rely on AI summaries and automated citation tools is significant. However, the Supreme Court has now made clear that such shortcuts are impermissible. Every precedent cited in an insolvency petition, response, or application must be verified against the actual judgment text. The NCLT and appellate courts will expect this level of diligence.

The Bar Council of India's forthcoming examination of professional conduct standards will likely result in specific rules about AI use in legal practice. These rules may require lawyers to disclose when they have used AI tools in research or drafting, to maintain records of verification steps taken, and to face liability if unvalidated AI output is submitted to courts. Some bar associations in other jurisdictions have already begun developing such rules, and India's Bar Council will likely draw on international experience in formulating its own guidelines.

The broader context is important. This judgment comes at a time when AI adoption in legal services is accelerating globally. Courts worldwide are grappling with similar issues. The Supreme Court of India's clear statement on this issue positions Indian courts as taking a firm stance on the authenticity and reliability of legal materials, which reinforces public confidence in the judicial system. If courts allowed decisions to rest on fabricated precedents, the entire edifice of precedent-based jurisprudence would be weakened.

For litigants in insolvency cases, the ruling provides a safeguard. It ensures that NCLT orders will not be based on fake law. If a lower tribunal has relied on unvalidated AI-generated precedents, appellate courts now have explicit authority from the Supreme Court to set aside such orders. This creates a mechanism for correcting errors that might otherwise go undetected.

The immediate effect of the judgment is that the NCLT order in Pooja Ramesh Singh's case has been set aside to the extent it relied on the fabricated precedents. The case will likely be remitted to the tribunal for reconsideration based on authentic legal authority. The Bar Council of India has been given a timeline, though the judgment does not specify a date, to report back to the Court on what steps it has taken to address professional conduct issues related to AI use.

For the legal profession, the message is clear: artificial intelligence is a tool, not a substitute for professional judgment and verification. Lawyers who use AI must treat its output with skepticism and verify everything before relying on it in court. The Supreme Court has made the stakes of non-compliance explicit: orders can be set aside, cases can be remitted, and professional conduct complaints can be filed. The zero-tolerance approach announced in this judgment will shape how the legal profession uses AI for years to come.