The Supreme Court of India on Thursday struck down decisions issued by two insolvency courts after discovering that the tribunal relied on fictitious and nonexistent judicial precedents created using artificial intelligence (AI).
The National Company Law Tribunal's (NCLT) and the National Company Law Appellate Tribunal's (NCLAT) judgments admitting the company into corporate insolvency were overturned by the Supreme Court, which remanded the case to the NCLT for a new, fact-based decision.
Essel Infraprojects Ltd. acted as the corporate guarantor for the loan received by another company, while Jammu & Kashmir Bank initiated insolvency proceedings against the company.
Pooja Ramesh Singh, who was the suspended director of Essel Infraprojects, had filed a Section 7 petition challenging the entry of the company into the corporate insolvency resolution process. For the first time, the NCLT Mumbai Bench admitted the insolvency petition on August 28, 2024. The NCLAT confirmed the same on September 11, 2025.
Senior counsel Madhavi Divan, who was representing Essel Infraprojects Ltd., brought the problem to the court's attention by pointing out that the NCLT had relied on several judgments that either did not exist or had paragraphs that were not present in the original judgments.
The Supreme Court's independent verification verified the problem. It discovered that while some citations mentioned actual Supreme Court rulings but added fake paragraphs attributed to the court, others linked to completely nonexistent judgments.
The use of phony, AI-generated papers must be treated with ‘zero tolerance,’ according to a bench made up of Justices Pamidighantam Sri Narasimha and Alok Aradhe. The bench further stated that the integrity of the legal system is seriously threatened by the use of such hallucinated data without adequate verification. The case, according to the bench, raises more general questions about the application of AI to legal decision-making.
The judges clearly differentiated between AI being employed as a tool and replacing human judgment by saying, “It is not only a device to help us in our work; it is an alternative to our very thoughts, reasoning, and even decision-making.” Nevertheless, it was made clear that the judges were not against the usage of AI in law practice or courts.
They emphasized that. “While it can indeed be a tool in the courtroom, it must be supervised completely by human beings.” Further, the court directed the Bar Council of India to form a committee to properly study the issue.
Errors made by AI have been pointed out by Indian courts in the past. Learning about a lower court referring to four fake case statutes generated by an AI software, the High Court of Andhra Pradesh issued a warning to trial court judges in January 2026.
Similarly, in 2025, the Bombay High Court found a tax assessment order citing fake court precedents and the Supreme Court found that there were completely fabricated citations in a filed reply.
Also Read: How to Write Better Prompts for High-Performing AI Agents
The Supreme Court published draft regulations in June 2026 in response to this expanding practice. The regulations strongly forbid the use of AI tools in judicial decision-making or case outcome determination, even while they allow them for administrative and research tasks like legal research, writing, translation, transcribing, and case administration.
The document specifically cautioned against the dangers of automated content's unreliability, erroneous citations, and AI hallucinations.