The Supreme Court of India has delivered a landmark judgment addressing the perils of Artificial Intelligence (AI) in adjudication, particularly the reliance on AI-generated fake and hallucinated material as legal precedents. In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., the Court set aside orders from the National Company Law Tribunal (NCLT) and the National Company Law Appellate Tribunal (NCLAT) that cited non-existent judgments, underscoring the need for absolute human oversight in judicial decision-making. The judgment reaffirms the sanctity of adjudication and mandates zero tolerance for unverified AI-generated content, while also directing the Bar Council of India to frame guidelines to prevent such occurrences.
Background & Facts
The Dispute
The appellant, Pooja Ramesh Singh, is a suspended director of Essel Infraprojects Ltd. (EIL), which acted as the corporate guarantor for Pan India Utilities Distribution Company Ltd. (PIUDCL). PIUDCL had availed loan facilities from Jammu and Kashmir Bank Ltd. (Respondent No. 1). Following defaults in repayment, the bank classified the loan accounts as non-performing assets and initiated corporate insolvency resolution proceedings under Section 7 of the Insolvency and Bankruptcy Code, 2016 (IBC) against EIL.
Key Facts
- 18.11.2017: Renewed sanction letter executed, allegedly omitting the guarantee clause.
- Financial Stress: PIUDCL defaulted on repayments, leading to classification as a non-performing asset.
- Guarantee Deed: Clause 8 stipulated that the guarantee would not be determined by amalgamation or absorption of the guarantor company.
Procedural History
- Section 7 Application: Filed by Jammu and Kashmir Bank Ltd. before the NCLT, Mumbai (RCP (IB) 6/MB/2023).
- 28.08.2024: NCLT admitted the application, appointed an Interim Resolution Professional, and declared a moratorium under Section 14 of the IBC.
- Appeal: The appellant challenged the NCLT order before the NCLAT (Company Appeal (AT) (Insolvency) No. 1808 of 2024).
- 11.09.2025: NCLAT dismissed the appeal, relying on judgments cited by the NCLT.
- Present Appeal: The appellant approached the Supreme Court, highlighting that the cited judgments were fake and AI-generated.
The Parties' Positions
Appellant’s Contention
- The NCLT erred in relying on non-existent judgments.
- The renewed sanction letter dated 18.11.2017 implied relinquishment of the guarantee.
- The scheme of demerger and amalgamation transferred liabilities, absolving EIL of its guarantee obligations.
Respondent’s Contention
- The guarantee deed explicitly stated that amalgamation would not affect the guarantee (Clause 8).
- The bank relied on the NCLT’s research, not submissions from its counsel.
Relief Sought
The appellant sought:
- Setting aside the NCLT and NCLAT orders.
- Restoration of the Section 7 application for fresh adjudication.
- Directions to the NCLT to decide the matter on merits without reliance on fake precedents.
The Legal Issue
The central question was whether judicial decisions based on AI-generated fake or hallucinated material can be sustained in law, and what consequences and accountability should follow for the Bar and Bench in such cases.
Arguments Presented
For the Appellant
- Ms. Madhavi Divan, learned senior counsel, argued that the judgments cited by the NCLT were non-existent and likely AI-generated.
- Even where citations were accurate, the excerpted paragraphs were untraceable in law reports.
- The reliance on such material vitiated the entire adjudication process.
For the Respondent
- The respondents countered on merits, arguing that the guarantee deed’s Clause 8 was binding.
- An affidavit was filed stating that the bank’s counsel did not cite the fake judgments; they were obtained through the NCLT’s own research.
The Court's Analysis
The Supreme Court undertook a rigorous examination of the implications of AI in adjudication, emphasizing the following principles:
-
Human Control Over AI: While AI can aid judicial processes, absolute human oversight is non-negotiable. The Court observed:
"Artificial Intelligence has acquired the capability to better, if not fully substitute, human effort... This extraordinary capability... is compelling professionals to adapt and employ AI for intelligent, efficient, and swift functioning. Yet if left unregulated, Artificial Intelligence may infiltrate our intellectual work ethic and render us dependent on its vast capabilities."
-
Sanctity of Adjudication: The Court declared that reliance on AI-generated fake material undermines the rule of law. It likened such material to "methyl isocyanate" - invisible, insidious, and catastrophic.
"It not only contaminates but takes away the very lifeblood of judicial determination."
-
Zero Tolerance Policy: The Court mandated zero tolerance for citing or relying on unverified AI-generated precedents, whether by the Bar or Bench. It held:
"Such a decision is no decision in the eyes of the law... It amounts to subversion of the rule of law."
-
Accountability and Regulation: The Court directed the Bar Council of India to constitute a committee to deliberate on disciplinary actions for advocates submitting fake AI-generated material. It also acknowledged the need for public policy and enforceable regulations to govern AI usage in legal practice.
-
Judicial Verification: The Court highlighted the impracticality of verifying every cited precedent but stressed that judicial trust in advocates’ submissions must not extend to unverified AI-generated content.
The Verdict
The Supreme Court set aside the NCLT order dated 28.08.2024 and the NCLAT order dated 11.09.2025. The Section 7 application (RCP (IB) 6/MB/2023) was restored to its original number, and the NCLT was directed to:
- Proceed with the application and decide it on merits.
- Dispose of the matter expeditiously, preferably within two weeks from the date of the judgment.
- Maintain status quo pending disposal.
The Court clarified that it had not expressed any opinion on the merits of the case, leaving the adjudicating authority to decide based on facts and law.
What This Means For Similar Cases
Zero Tolerance for AI-Generated Fake Precedents
- For Advocates: Citing AI-generated fake judgments constitutes professional misconduct. The Bar Council of India is now tasked with framing guidelines and disciplinary actions for violations.
- For Judges: Reliance on unverified AI material will render judgments unsustainable and liable to be set aside, even if the fake material had only an indirect bearing on the decision.
Human Oversight Remains Non-Negotiable
- AI can assist in research and drafting, but final verification and decision-making must remain human-centric. The Court emphasized that delegating thinking to AI erodes the core of judicial reasoning.
- Practitioners must cross-verify AI-generated content against authoritative legal texts before submission.
Need for Regulatory Frameworks
- The judgment underscores the urgent need for enforceable regulations governing AI usage in legal practice. While the Court acknowledged ongoing deliberations, it stressed that self-regulation by the Bar and Bench is critical.
- Public policy must address the risks of AI hallucinations, ensuring that technological advancements do not compromise judicial integrity.
Precedential Value and Global Implications
- The judgment serves as a warning against over-reliance on AI in jurisdictions worldwide. It aligns with international incidents, such as the UK case of Anthony Malcolm Cork v. Mark Smith, where a law firm’s reckless use of AI led to regulatory scrutiny.
- Courts globally may now adopt stricter scrutiny of AI-generated submissions, reinforcing the principle that judicial decisions must be based on verified, authentic material.
