AI-Hallucinated Precedents Render Judicial Decisions Void: Zero Tolerance in Adjudication

Introduction

In POOJA RAMESH SINGH v. JAMMU AND KASHMIR BANK LTD., the Supreme Court of India addressed a grave procedural and institutional concern: reliance by a tribunal on fake, non-existent, or AI-generated “hallucinated” legal precedents.

The dispute arose from insolvency proceedings under Section 7 of the Insolvency and Bankruptcy Code, 2016. Jammu and Kashmir Bank Ltd. had initiated corporate insolvency resolution proceedings against Essel Infraprojects Ltd., a corporate guarantor for loans advanced to Pan India Utilities Distribution Company Ltd. The NCLT admitted the application, and the NCLAT affirmed that decision.

The appellant, a suspended director of the corporate debtor, challenged the orders before the Supreme Court, contending that the NCLT had relied on fake and non-existent judgments, apparently generated through Artificial Intelligence.

Summary of the Judgment

The Supreme Court set aside both the NCLT order dated 28.08.2024 and the NCLAT judgment dated 11.09.2025. The Court restored the Section 7 application to the NCLT for fresh consideration on merits.

The Court held that any judicial or quasi-judicial decision relying on fake, hallucinated, or non-existent precedents is no decision in the eyes of law. It declared a rule of zero tolerance for the use of AI-generated legal material without verification.

Importantly, the Court clarified that it was not condemning the proper use of AI in legal work. Rather, it condemned the presentation or reliance upon hallucinated material as if it were binding or persuasive legal precedent.

The Supreme Court also directed the Bar Council of India to constitute a committee and frame guiding principles and disciplinary consequences for advocates who submit fake or hallucinated legal material before courts.

Analysis

Precedents Cited

The judgment is significant because the “precedents” relied upon by the NCLT were themselves under scrutiny. The Supreme Court independently examined them and found serious defects.

  • State Bank of India Vs. M/s Shree Ram Urban Infrastructure Ltd.
    This was cited as 2020 SCC OnLine SC 341. The Supreme Court found that the citation actually corresponded to another case, namely M. Subramaniam v. S. Janaki. The paragraph attributed to the cited case was also non-existent. Thus, the tribunal had relied on a wrongly attributed and fabricated legal proposition.
  • Everest Kento Cylinders Ltd. Vs. Union of India
    The citation (2015) 2 SCC 1 was found to be a correct citation, but the paragraph attributed to the judgment was not traceable. This showed that even genuine citations can be misused by attaching fabricated extracts to them.
  • ICICI Bank Ltd. Vs. Urban Infrastructure Real Estate Ltd.
    The Supreme Court found this citation to be non-existent. It was therefore not a valid legal authority at all.
  • V.S Dempo & Co. Ltd. Vs. Reliance Communications Ltd.
    This too was found to be a non-existent citation. Its use demonstrated the danger of relying on unverified AI-generated outputs.
  • Canara Bank Vs. N.G. Subbaraya Setty & Anr.
    The citation was correct, but the paragraph relied upon was non-existent. The Court treated this as equally problematic because fabricated reasoning cannot be legitimised merely by attaching it to a real case name.
  • Sarbjit Singh Vs. Union Bank of India
    This was found to be a non-existent citation. It therefore could not have formed part of any lawful adjudicatory reasoning.
  • Anthony Malcolm Cork & Anor v. Mark Smith, [2026] EWHC 1199 (Ch)
    The Supreme Court referred to this English decision to highlight similar risks arising from careless use of AI in legal practice. In that case, a law firm had relied on fabricated statutory text generated by AI. Though contempt proceedings were not initiated due to absence of deliberate dishonesty, the incident showed how AI misuse can waste judicial time and damage professional standards.

Legal Reasoning

The Court’s reasoning rests on the integrity of adjudication. Legal decisions must be based on real statutes, real precedents, verified facts, and reasoned application of law. If a court relies on non-existent authorities, the process is contaminated.

The Supreme Court described AI hallucinated precedents as an invisible and catastrophic threat to justice. According to the Court, even a small amount of fake material entering the decision-making process is sufficient to vitiate the decision.

The Court laid down three core principles:

  1. AI may assist adjudication, but cannot replace human judicial responsibility.
  2. Advocates commit misconduct if they cite AI-generated material without verification.
  3. A judicial decision relying on fake or hallucinated precedents is no decision in law.

Applying these principles, the Court found that the NCLT’s order was tainted. The NCLAT also failed to detect the fake precedents. Therefore, both orders had to be set aside, regardless of the merits of the insolvency claim.

Impact

This judgment is likely to become a leading authority on the use of Artificial Intelligence in legal proceedings. Its immediate impact is procedural: courts and tribunals must ensure that cited authorities are real and accurately represented.

For advocates, the judgment creates a professional duty of verification. Lawyers cannot rely blindly on AI tools, databases, or generated summaries. If they submit fake case law, disciplinary consequences may follow.

For judges and tribunals, the judgment imposes a duty of caution. While courts may use technology, final responsibility for reasoning and decision-making must remain human.

In insolvency law specifically, the judgment does not decide the substantive liability of the corporate guarantor. Instead, it restores the Section 7 application for fresh adjudication. However, its broader effect extends far beyond insolvency: it protects the rule of law from fabricated legal authorities.

Complex Concepts Simplified

  • AI hallucination: A situation where an AI system generates information that sounds authoritative but is false, such as a fake case citation or fabricated paragraph from a judgment.
  • Precedent: A previous judicial decision that courts may follow or consider while deciding later cases.
  • Section 7 of the IBC: A provision allowing a financial creditor to initiate insolvency proceedings against a corporate debtor when there is debt and default.
  • Corporate guarantor: A company that promises to repay a debt if the principal borrower defaults.
  • Moratorium: A temporary legal pause on certain proceedings and recovery actions once insolvency proceedings are admitted.
  • Human in the loop: The principle that AI may assist, but a human professional must verify, supervise, and remain accountable for the final legal work.

Conclusion

The Supreme Court’s decision establishes a clear and important rule: courts cannot tolerate fake or AI-hallucinated precedents in adjudication. Any decision tainted by such material is legally unsustainable.

The judgment balances technological openness with institutional caution. It recognises AI as a useful tool, but insists that legal reasoning, professional responsibility, and judicial accountability must remain firmly human.

The key takeaway is simple: AI may assist the law, but it cannot fabricate the law.