Buried in the record of an Indian insolvency dispute sat case law that no judge had ever written. When India’s Supreme Court examined the orders passed by the National Company Law Tribunal and its appellate body, the NCLAT, it found citations to judgments that simply do not exist. The court set both orders aside. The fabricated precedents, most likely the product of AI hallucinations, have turned a routine corporate dispute into a warning shot for the legal profession.
How fake law reaches real judges
Generative AI systems are built to predict the next likely sequence of words, not to distinguish an actual Supreme Court precedent from one they have inadvertently invented. Conventional legal databases index authenticated judgments. A chatbot produces text that looks like one. When lawyers, and increasingly judicial officers, use these tools to summarise judgments, draft pleadings and speed up research, convincing fabrications can slip into filings unless a human checks every citation against the source.
Hallucination is the industry’s term for this failure, and it is not a bug that better prompting fixes. A model generating a plausible-sounding citation, complete with party names, a year and a paragraph of confident reasoning, is doing exactly what it was designed to do. The confidence is the danger. A fabricated case does not arrive flagged as uncertain. It arrives looking like law.
That checking clearly did not happen here, and the failure ran through multiple layers. A tribunal decided a dispute on authority that never existed, an appellate tribunal let it stand, and only India’s highest court caught it.
The accountability gap
The uncomfortable question raised by the ruling is who answers for a fabricated citation once it enters the record. The lawyer who used the tool, the judge who relied on the research, the tribunal that reproduced it, or the company that built the model? Indian courts currently offer no framework for the answer. There are no judicial rules requiring advocates to disclose when AI helped prepare pleadings, and no prescribed standards for verifying AI-generated research before it is cited.
The absence of rules has not slowed adoption. Lawyers, law firms and judicial officers across India are experimenting with AI to compress weeks of research into hours, and the efficiency gains are real. That is precisely what makes the risk structural rather than incidental: the tools are spreading faster than the verification habits needed to use them safely.
Saakar Yadav, founder of legal research firm Lexlegis AI, argues the vulnerability is universal. Any court, firm or regulator that entrusts research to systems optimising for plausibility rather than truth is exposed, he told THE WEEK, and the future of legal AI will depend less on larger models than on verified legal corpora that keep machine reasoning anchored to authentic sources of law.
Other countries are further along
In the United States, multiple lawyers have already faced sanctions for filing briefs containing fictitious AI-generated precedents. Some federal judges now require counsel to certify that every citation has been independently verified and that nothing AI-generated has been filed without human review. Canada, Australia and the United Kingdom are having similar debates. India, despite rapid AI adoption across its legal profession, has yet to frame comparable safeguards.
American courts are also wrestling with subtler problems than fake citations. In the first quarter of 2026, federal courts issued diverging decisions on whether feeding case material into public AI tools waives legal protections. A court in the Southern District of New York held that using public generative AI tools could destroy privilege protections in a criminal matter, while courts in Michigan and Colorado found that civil litigants did not waive work product protections merely by using them. The ground rules are being written case by case.
THE WEEK reports that legal experts expect Indian courts to eventually consider practice directions requiring mandatory verification of AI-generated citations, disclosure of AI-assisted drafting in appropriate cases, and regular training for judges and lawyers on what these tools can and cannot do.
Courts derive their legitimacy from getting the law right. If invented authorities begin steering judicial reasoning, even occasionally, public confidence erodes in a way that is hard to rebuild. The Supreme Court has now put every tribunal in the country on notice, and the next fabricated citation that surfaces in an Indian courtroom will land in a very different climate. For more coverage of AI and the law, visit Mylistingo.
