Artificial Intelligence in Legal Research: Judicial Guidance, Misuse, and Compliance Duties

Artificial Intelligence has rapidly moved from being a futuristic concept to a pervasive tool across professions, including the legal sector. Legal research, drafting, and analysis are now frequently supported by AI-driven platforms and Large Language Models (LLMs). While these tools can substantially enhance efficiency and provide useful preliminary insights, courts in India and abroad have increasingly been confronted with a troubling trend: unverified or fabricated case law being placed before judicial and quasi-judicial authorities, often traceable to AI-generated content.

This article distils how various courts – particularly in India, the United Kingdom, and Canada – have responded to the misuse of Artificial Intelligence in legal proceedings, especially in the context of legal research and precedent citation. It also outlines the professional responsibilities of advocates, quasi-judicial authorities, and regulators in this evolving environment.

1. AI-Generated “Judgments” Before Indian Courts

1.1 Supreme Court’s intervention in GUMMADI USHA RANI & ANR. Vs SURE MALLIKARJUNA RAO & ANR.

In GUMMADI USHA RANI & ANR. Vs SURE MALLIKARJUNA RAO & ANR. (Petition for Special Leave to Appeal (C) No. 7575/2026, heard on 27.02.2026), the Supreme Court was confronted with a serious issue: a Trial Court had relied on judgments that appeared to be AI-generated and not genuine reports of decided cases.

The matter arose in the context of objections raised by defendants in a civil suit. While the High Court, in revision, had issued a cautionary note and nevertheless decided the matter on merits, the Supreme Court took a far stricter view. It observed that the Trial Court had seemingly used “AI generated non-existing, fake or synthetic alleged judgments” and categorically held that basing a decision on such material was not merely an error in adjudication, but misconduct. The Court clarified that:

A decision founded on non-existent or fabricated precedents is not simply a wrong decision; it is a serious breach affecting the integrity of the adjudicatory process, warranting legal consequences.

Recognising the systemic implications, the Supreme Court issued notices to the Ld. Attorney General, Ld. Solicitor General and the Bar Council of India to examine the broader consequences and accountability mechanisms relating to the use of AI-generated legal materials.

1.2 Gujarat High Court’s concerns in Marhabba Overseas Pvt. Ltd. Vs Union of India

In Marhabba Overseas Pvt. Ltd. Vs Union of India (decided on 20-2-2026), the Division Bench of the Gujarat High Court grappled with quasi-judicial authorities relying either on:

  • Judgments that did not exist at all, or
  • Real judgments that had no connection with the controversy before them.

The Court found this to be an appropriate case to lay down parameters for quasi-judicial authorities when citing judgments of the High Courts or the Supreme Court of India. It emphasised that when dealing with legal issues raised by an assessee, reliance on precedent must be:

  • Genuine,
  • Relevant, and
  • Duly verified.

The underlying concern was that with AI-assisted research, authorities might be tempted to rely on unverified outputs, which can corrupt the quality and legitimacy of adjudication.

2. AI Usage in Tax and GST Proceedings

2.1 Delhi High Court in J.M. Jain Vs Union of India

In J.M. Jain Vs Union of India (decided on 27-11-2025), the Division Bench of the Delhi High Court was dealing with a Show Cause Notice (SCN) under GST that followed investigations carried out under the Income Tax law. The Revenue cited several judicial precedents to support its position. On closer examination, certain anomalies and inaccuracies in the cited judgments came to light.

The Court acknowledged that:

  • Government departments are free to use Artificial Intelligence tools to analyse evidence, prepare summaries, and assist in research,
  • Such use is permissible only subject to strict verification of the results.

However, the Court stated in unmistakable terms that the Departments – including GST and Income Tax authorities – cannot rely on fake or non-existent judgments. The fact that such judgments may have been surfaced or formatted by AI tools was no defence.

The Delhi High Court noted that various High Courts had already cautioned against the uncritical use of Artificial Intelligence for case law citation and stressed that:

Before issuing SCNs or finalising assessments, all judgments relied upon must be independently verified by the authorities, regardless of whether they were initially obtained via AI software.

2.2 Bombay High Court in KMG Wires (P.) Ltd. v. National Faceless Assessment Centre

A similar issue arose before a Coordinate Bench of the Bombay High Court in KMG Wires (P.) Ltd. v. National Faceless Assessment Centre (decided on 06.10.2025). The challenge was to an assessment order under Section 143(3) read with Section 144B of the Income Tax Act 1961.