A judge in Andhra Pradesh once based an order on four judgments that never existed. In Bengaluru, a tax tribunal recalled a ₹669-crore order after discovering it rested on citations nobody could find. In Mumbai, the High Court fined a litigant ₹50,000 for a filing so obviously AI-written it still carried the chatbot's green tick marks.
None of these are edge cases anymore. They are the new normal, and courts everywhere are done tolerating it.
An AI hallucination happens when a tool like ChatGPT invents a case citation, a quote, or an entire judgment with total confidence, and none of it is real. Large language models are built to predict plausible-sounding text, not to retrieve verified legal records. The result reads exactly like a real precedent, right down to the case name, the court, and the year. The only problem is that it never happened.
Lawyers who skip the one step that matters, actually checking the citation against a real database, are the ones who end up in front of a disciplinary bench.
On July 2, 2026, the Supreme Court of India set aside NCLT and NCLAT orders in the Essel Infraprojects insolvency matter after finding that six of the judgments relied on were defective. Three did not exist at all. The other three were real cases stuffed with invented paragraphs or wrong titles.
The Court did not treat this as a technical slip. It called hallucinated case law comparable to a toxic gas leak in the province of law and justice, directed the Bar Council of India to frame disciplinary norms, and made clear that a decision resting even partly on fake material is no decision at all in the eyes of the law.
This built on an earlier warning. In February 2026, the Court took suo motu notice of a trial court order in the Gummadi Usha Rani matter that had relied on fabricated judgments, and declared that this is not an error in the decision-making process. It is misconduct, with legal consequences that follow.
The Essel ruling did not come out of nowhere. It is the latest entry in a growing list of Indian cases where fake AI citations blew up in court.
Bombay High Court, January 2026. A litigant was fined ₹50,000 for dumping AI-generated case laws into written submissions, complete with the telltale formatting of raw chatbot output.
Delhi High Court, September 2025. A petition was withdrawn in embarrassment after opposing counsel exposed the citations as fabricated, including invented paragraphs attributed to a landmark constitutional case.
Punjab and Haryana High Court, early 2026. Judges were directed not to rely on AI while drafting judgments, on the reasoning that technology cannot substitute for human reasoning.
Income tax and GST tribunals, 2025 - 2026. Multiple orders, including a roughly ₹28-crore assessment upheld by the Bombay High Court, were quashed after AI-invented precedents surfaced in the record. Quasi-judicial officers, not just lawyers, have been found using AI-fabricated content in official proceedings.
The same story is playing out abroad. In the United States, a public tracker has now logged more than 1,600 court decisions worldwide involving AI-hallucinated material. Sanctions have climbed from a symbolic 5,000 dollar fine in 2023 to more than 110,000 dollars in a single Oregon matter, and an Omaha attorney was suspended in April 2026 after denying, then admitting, that AI had fabricated the bulk of his citations. England and Wales, Australia, South Africa, and Argentina have all recorded similar incidents.
Across nearly every sanctioned case, one detail repeats. Courts punish concealment far more harshly than the original error.
Lawyers who admitted the mistake immediately and corrected the record generally received lighter consequences. Lawyers who denied using AI, blamed a junior associate, or quietly refiled without disclosure were the ones who lost their license to practice or faced six-figure penalties. Candor is the one thing a chatbot cannot supply on a lawyer's behalf.
The judicial response is no longer limited to warnings. It now includes:
Setting aside any order that relies even partly on unverified AI material, regardless of the outcome's substantive merit.
Directing bar councils to frame formal disciplinary norms specifically for AI misuse.
Escalating monetary sanctions well beyond token fines, with some orders now calculated per fake citation and per fabricated quote.
Referring repeat or dishonest offenders for suspension.
Instructing judges and tribunal members themselves not to rely on AI-drafted material without independent verification.
The direction of travel is unmistakable. Zero tolerance is no longer a phrase in a judgment. It is becoming the operating standard.
If you are a litigant, an order built on a hallucinated citation may be challengeable on that ground alone, even if you never raised the underlying facts. If you are a lawyer or in-house counsel, every AI-assisted draft now needs the same verification discipline as any other piece of work product, because supervision liability falls on the person who signs the filing, not the tool that drafted it. If you are relying on a tribunal order that turns out to rest on fake precedent, you may have real grounds to have it recalled.
Can a court order be challenged just because it cites a fake case?
Yes. Indian courts have already set aside and recalled orders solely because the citations relied upon did not exist or were misattributed, independent of the case's other merits.
Is using AI to draft legal documents banned in India?
No. The Supreme Court's own white paper on AI in the judiciary endorses its use as a research aid. What is prohibited is filing or relying on AI output without independently verifying every citation.
Who is held responsible when AI generates a fake citation, the lawyer or the AI tool?
The lawyer. Courts have consistently held that the duty of verification and candor sits with the advocate who signs and files the document, regardless of which tool produced the draft.
Can a client be penalized for their lawyer's AI mistake?
Directly, sanctions are usually imposed on counsel. Indirectly, clients have faced real damage, including opposing counsel's fees that were not reversed even after their own lawyer was sanctioned, and orders that had to be relitigated from scratch.
What should I do if I suspect an order against me relied on a fake precedent?
Get the citations independently verified against an authentic legal database before you file anything. If a fabrication is confirmed, it can form an independent ground for appeal, review, or recall.
If an order, assessment, or tribunal ruling against you was built on a citation that turns out not to exist, you are not without a remedy, and you do not have to work out the legal pathway on your own.
Fairaigle Legal & Consultancy LLP works with individuals, businesses, and professionals who have been affected by AI-hallucinated case law, whether the fabrication came from the other side's filing, a tribunal's own order, or an assessment that relied on invented precedent. Our team independently verifies every citation in the record against authentic sources, identifies the strongest ground for appeal, review, or recall, and represents you through the process of getting a flawed order set aside.
We also help lawyers and businesses build the kind of verification workflow that keeps them out of this problem entirely, so an AI-assisted draft never becomes a liability with your name on it.
If you believe you have been affected by fake case law in your own matter, do not wait for the next hearing to find out the hard way.
Talk to Fairaigle Legal & Consultancy LLP today for a confidential case review.
Event Images not Available
