The Supreme Court of India has drawn a firm line against the misuse of artificial intelligence in litigation. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr. (2026 INSC 668; reported as 2026 LiveLaw (SC) 653), a Bench of Justices P.S. Narasimha and Alok Aradhe, deciding the matter on 2 July 2026, held that reliance on AI-generated fake or "hallucinated" case law strikes at the very integrity of the judicial process. The Court set aside orders passed by the National Company Law Tribunal (NCLT) and the National Company Law Appellate Tribunal (NCLAT) that had rested on fabricated precedents, and directed the Bar Council of India to frame guidelines governing the use of AI-generated material in legal proceedings. This judgment is likely to become the reference point for how Indian courts treat generative AI in the years ahead.
Background: how fabricated precedents entered the record
The dispute arose out of insolvency proceedings. An application had been filed under Section 7 of the Insolvency and Bankruptcy Code, 2016 (IBC) in connection with Essel Infraprojects, seeking to initiate the corporate insolvency resolution process against the corporate debtor. During the litigation before the NCLT and, subsequently, the NCLAT, submissions were made that relied on judicial precedents said to support the parties' positions.
The difficulty was that several of those precedents did not exist. Some "citations" referred to judgments that had never been delivered by any court. In other instances, genuine case citations had been padded with paragraphs and observations that were never part of the actual judgment — text that had, in effect, been generated by an AI tool and passed off as authoritative reasoning. These fabricated authorities found their way into the orders of the tribunals, which relied on them while adjudicating the rights of the parties.
When the matter travelled to the Supreme Court, the fabrication came to light. The Court was confronted not merely with an error of law but with something more corrosive: decisions that had been built, at least in part, on precedents that were pure invention.
The central question before the Court
The core issue was not whether AI may be used by lawyers at all. Generative tools are now a routine part of legal research and drafting. The question, rather, was what happens to the sanctity of adjudication when unverified, machine-generated material is presented to a court as genuine precedent — and what responsibility rests on the advocate who places such material before the Bench.
The Court framed the problem as one going to the foundation of the justice system. Precedent is the currency of common law adjudication. If that currency can be counterfeited — if judges and tribunals can be induced to decide cases on the strength of authorities that were never authored by any court — then the reliability of every judgment is placed in doubt.
Key holdings
The Bench's reasoning can be distilled into a few propositions that practitioners should note carefully.
Zero tolerance for hallucinated authority. The Court held that reliance on hallucinated material strikes at the integrity of adjudication and its processes, and that courts must adopt a zero-tolerance approach when confronted with fake AI-generated precedents produced by lawyers without verification. The standard, in other words, is not one of leniency for an honest mistake with new technology; it is one of strict responsibility.
Citing unverified AI precedents can amount to professional misconduct. Significantly, the Court did not treat the problem as a mere procedural irregularity. Placing fabricated authorities before a court, without having verified them against the official record, was characterised as conduct capable of attracting the discipline applicable to professional misconduct. The duty of candour that an advocate owes to the court is not diluted merely because a machine, rather than the advocate, generated the offending text.
The orders below could not stand. Because the impugned NCLT and NCLAT orders were founded on unverified AI-generated authorities, the Supreme Court set them aside. Reliance on such material, the Court reasoned, fundamentally undermines judicial integrity and cannot be allowed to support a binding adjudication.
Directions issued
Recognising that individual censure would not be enough to address a systemic risk, the Court looked to institutional safeguards. It directed the Bar Council of India to constitute a committee to frame guidelines and to consider appropriate disciplinary measures governing the use of AI-generated material in legal proceedings. The object is to create a clear, enforceable framework — so that the profession is not left to work out its obligations case by case, and so that the consequences of placing unverified material before a court are known in advance.
This direction dovetails with a broader institutional movement. The judgment arrives alongside the framing of "Regulations for the Use of Artificial Intelligence in Courts, 2026," reflecting a wider effort within the justice system to set boundaries for AI use rather than to prohibit it outright. The message from the Court is calibrated: technology is welcome as an aid, but it cannot be permitted to displace the professional judgment and verification that lie at the heart of legal practice.
Why this judgment matters
For the practising Bar, the ruling removes any lingering ambiguity about where responsibility lies. The convenience of generative AI does not transfer the risk of error away from the advocate. If a citation is placed before a court, the lawyer is answerable for its authenticity, full stop. "The software produced it" is not a defence; it is, if anything, an admission that the necessary verification was not done.
For litigants, the decision is reassuring. It affirms that the courts will police the integrity of the material on which they act, and that a party cannot gain an advantage — or suffer a loss — on the strength of invented precedent. For the tribunals, it is a reminder that the duty to test the authorities cited does not evaporate merely because both sides are represented by counsel.
More broadly, the judgment situates India within a global conversation. Courts in other jurisdictions have already sanctioned lawyers for filing briefs containing AI-fabricated citations. By articulating a zero-tolerance standard and directing the Bar Council to build a regulatory framework, the Supreme Court has moved India from ad hoc responses toward a structured approach.
Practical takeaways for lawyers
The safest reading of this judgment is that verification is now a non-negotiable professional obligation whenever AI is used in research or drafting. A few disciplines follow naturally:
- Verify every citation against a primary source. Before a case is cited, confirm that it exists and that the proposition attributed to it actually appears in the reported judgment — ideally on an official database or the court's own record, not merely in an AI summary.
- Treat AI output as a draft, not an authority. Generative tools are useful for structuring arguments and surfacing lines of inquiry, but their output must be checked line by line before it is relied upon.
- Be wary of "padded" precedents. Even genuine citations can be corrupted by fabricated paragraphs. Read the actual judgment, not a machine-generated paraphrase of it.
- Keep a verification trail. Maintaining a record of how authorities were checked can protect an advocate if the authenticity of a citation is ever questioned.
Conclusion
The decision in Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. is less a case about insolvency than a case about trust — the trust that a court must be able to place in the material laid before it. By insisting on a zero-tolerance standard, treating unverified AI citations as a matter of professional responsibility, and directing the Bar Council of India to build a lasting framework, the Supreme Court has adapted a very old principle to a very new tool: an advocate is, and remains, personally accountable for the truth of what they tell the court. As AI becomes further embedded in legal practice, that principle is likely to be tested again — but after this judgment, no one can plausibly claim they were unaware of it.
This article is intended for general information and does not constitute legal advice. Readers should refer to the full text of the judgment and consult a qualified advocate for advice on any specific matter.
