Authored By: Aditi Bajaj
Chandigarh University
INTRODUCTION
The growing incorporation of generative artificial intelligence into legal research has fundamentally transformed the manner in which lawyers locate, analyse, and present judicial precedents. While these technologies offer significant gains in efficiency and accessibility, they also introduce unprecedented risks to the integrity of judicial decision-making. In early 2026, these risks materialised dramatically before the Supreme Court of India in Gummadi Usha Rani v. Sure Mallikarjuna Rao, where, upon a detailed examination of the trial court record, the Court discovered that the subordinate court had relied upon four entirely fictitious Supreme Court precedents while deciding a property dispute.[1] It was subsequently revealed that the legal representatives had relied on unverified generative artificial intelligence tools to prepare their submissions. Operating through predictive language generation rather than authenticated legal databases, the AI system fabricated realistic case names, citations, and judicial reasoning that bore no connection to actual judicial decisions.[2] The incident exposed a significant procedural vulnerability within the Indian legal system, which is that existing procedural safeguards governing the citation and reliance upon judicial authorities were developed to address human error, negligence, or deliberate misrepresentation, not a technological environment in which artificial intelligence can autonomously generate convincing but entirely fictional legal authorities.
This phenomenon is not confined to India. The decision in Mata v. Avianca, Inc.[3] similarly demonstrated that AI systems are capable of producing fabricated judicial authorities that appear authentic, resulting in false case citations being presented before courts. These incidents illustrate that AI hallucinations are no longer isolated technological failures but represent an emerging procedural challenge capable of undermining judicial accuracy, professional accountability, and public confidence in the administration of justice.
Accordingly, this article contends that although the existing procedural and ethical framework continues to provide an important foundation for safeguarding judicial integrity, it must be strengthened through a strict, non-delegable human verification requirement and structured procedural safeguards before AI-generated authorities can be relied upon in judicial proceedings.
EXISTING PROCEDURAL AND PROFESSIONAL SAFEGUARDS: ARE THEY SUFFICIENT?
The appearance of AI-created fictional judicial bodies does not automatically suggest that current legal systems are fundamentally flawed. Instead, it demands a careful evaluation of whether existing procedural and professional safeguards are sufficient to manage this emerging risk, considering the context of ongoing technological advancements. This section contends that even though Indian law does not explicitly govern AI-assisted legal research, its established evidentiary standards, procedural duties, and professional ethics together offer a robust framework for safeguarding judicial integrity.
Statutory Framework on Evidence and Judicial Notice
The core statutory framework regulating the admissibility of legal authorities in Indian courts is established under the recently passed Bharatiya Sakshya Adhiniyam, 2023. Under Section 56 of the Bharatiya Sakshya Adhiniyam, 2023,[4] facts that a court is required to take judicial notice of all laws in force within the territory of India, encompassing public Acts enacted by Parliament and the historical development of common law jurisprudence.
Indian evidence law mainly governs the acceptability, genuineness, and trustworthiness of documentary and electronic evidence. The Bharatiya Sakshya Adhiniyam, 2023 updates evidentiary rules by acknowledging electronic records and setting standards for their admissibility and authentication. These provisions guarantee that electronic documents submitted in court meet specified legal criteria before being accepted as evidence. However, the statutory framework is geared toward establishing facts rather than confirming legal authorities. A judicial precedent referenced in written submissions is not presented as documentary evidence but as a legal statement meant to inform judicial reasoning. Generative AI has fundamentally transformed this environment by allowing the rapid generation of convincing yet completely fabricated judicial authorities. The challenge stems from technological change rather than from insufficient legislation. The established rules of evidence still apply; what has shifted is the practical need to confirm authorities before they affect judicial reasoning.
When an advocate cites a case law precedent in oral or written submissions, they are essentially asking the court to treat that legal principle as judicially noticed under Article 141 of the Constitution of India, which mandates that the law declared by the Supreme Court must be binding on all courts.[5] The statutory framework assumes that a cited case constitutes an existing, verifiable public record. The law includes clear procedures for addressing human misinterpretations or conflicting readings of a precedent, yet it overlooks a significant gap concerning automated fabrication that replicates the precise linguistic tone of authentic Judicial text.
Professional Standards and Ethical Responsibilities
Legal practitioners in India are bound by ethical obligations outlined in the Bar Council of India Rules, which are established under the Advocates Act, 1961. Chapter II of Part VI of the Bar Council of India Rules outlines the stringent professional conduct and etiquette standards required of advocates.[6] Section I of this chapter describes an advocate’s main obligation to the court, requiring that a practitioner show respect, behave with complete dignity, and use only fair and honourable methods.[7]
In fact, as the use of AI-assisted legal research grows, advocates are under a greater obligation to ensure the accuracy and authenticity of every authority they include in pleadings or oral arguments. Assigning legal research to AI does not equate to shifting professional responsibility. The advocate remains responsible for the content of documents filed with the court, regardless of the technology used in their preparation. This principle has been clearly acknowledged in other legal systems. Likewise, the American Bar Association, in Formal Opinion 512 (2024), noted that lawyers employing generative AI remain accountable for ensuring the accuracy of all legal work produced with AI assistance and must independently verify legal authorities before relying on them.[8] These developments underscore a principle also consistent with Indian professional ethics: while technology may aid legal research, professional judgment cannot be delegated.
Contempt Jurisdiction
When an advocate’s behaviour exceeds ordinary negligence and risks undermining the fair administration of justice, courts invoke their inherent contempt power. Section 2(c) of the Contempt of Courts Act, 1971 defines criminal contempt as the act of publishing any material or performing any action that scandalises, prejudices, or significantly disrupts the proper conduct of any judicial proceeding or the administration of justice.[9] Traditionally, proving criminal contempt requires demonstrating either intent or reckless disregard for the sanctity of the judicial process.
Existing Safeguards: Adequate but Procedurally Incomplete
The above analysis shows that Indian law already includes strong protections able to tackle many of the risks tied to AI-generated false judicial authorities. Evidentiary law safeguards the authenticity of documentary evidence, professional ethics require competence and honesty, and procedural rules support rigorous adversarial review. None of these safeguards has grown outdated due to generative AI. Nevertheless, they all share a common limitation: each assumes that the judicial authorities referenced before the court actually exist. However, in the age of generative AI, this assumption needs to be reevaluated.
III. THE PROCEDURAL GAP EXPOSED BY GENERATIVE AI
The rise of generative artificial intelligence has presented a new challenge to legal adjudication by eroding a core assumption of judicial procedure: that authorities referenced in court actually exist. Historically, courts have depended on advocates to cite genuine legal authorities grounded in professional ethics, procedural rules, and the adversarial system. Though these safeguards were originally designed for a time when legal research was conducted by humans, the growing reliance on AI-assisted research has changed the underlying assumption. Modern generative AI systems are capable of generating judicial citations that mimic the style, structure, and wording of real case law, even though they are not grounded in any established legal database.
The resulting challenge is not just technological but institutional, since current procedural frameworks were never intended to validate authorities that can be instantly forged using artificial intelligence. Understanding this challenge hinges on distinguishing between evidentiary authenticity and the authenticity of judicial authorities. Evidence law mainly deals with whether documents, electronic records, and other materials presented in court are admissible, reliable, and properly authenticated. Under the Bharatiya Sakshya Adhiniyam, 2023, documentary and electronic evidence must meet specified criteria before it can be accepted in court proceedings.
These provisions guarantee that the material courts depend on is authentic and verifiable. However, they do not directly oversee the independent verification of judicial authorities referenced in legal filings. Judicial precedents are generally regarded as expressions of law rather than as evidence that needs to be authenticated. Therefore, procedural practice presumes that advocates rely solely on genuine and verifiable authorities. The ruling in Mata v. Avianca, Inc. shows how generative AI undermines this assumption.[10] In that instance, counsel depended on legal research produced by an AI chatbot, which created several fictional judicial decisions that never existed. The Mata decision shows that the main institutional flaw does not reside in evidence law but in the procedural assumptions that shape legal advocacy. Courts generally assume that lawyers fulfil their professional duties by independently verifying the sources they depend on. That presumption worked well when legal research relied on recognised reporters, official databases, and established legal publications. Current legal systems govern the validity of evidence and set ethical duties for advocates, yet they still rely on the unspoken assumption that referenced judicial authorities are inherently genuine. That assumption is no longer viable in an age when artificial intelligence can produce legally convincing yet completely fabricated precedents in seconds. Thus, the challenge calls for procedural adjustment rather than a complete overhaul of evidentiary doctrine.
COMPARATIVE PERSPECTIVES ON PROCEDURAL RESPONSES TO AI-GENERATED FICTITIOUS JUDICIAL AUTHORITIES
Other jurisdictions’ experiences show that legal responses to AI-generated fake judicial authorities have focused on reinforcing current professional and procedural duties, rather than establishing new legislative structures. Though the methods vary in form, they all embody a shared principle: employing artificial intelligence does not relieve a lawyer of the duty to verify that every authority submitted to a court is genuine and correctly reflects the law. In the United States, professional regulations now confirm that attorneys are still personally accountable for the competence, accuracy, and reliability of all legal work they produce with the help of generative artificial intelligence. In Formal Opinion 512 ( 2024), the American Bar Association stresses that AI-supported legal research must be supplemented by independent confirmation of legal sources, and that technology cannot substitute for professional legal judgment.[11]
A similar position has arisen in the United Kingdom. Legal practitioners are required by guidance from the Courts and Tribunals Judiciary and the Law Society of England and Wales to confirm the authenticity of authorities before using them in litigation. Instead of creating a distinct regulatory framework for artificial word, the guidance treats AI as a research tool that continues to be governed by existing professional obligations of competence, diligence, and candour toward the court.[12] While India has not yet implemented comparable professional standards specifically targeting AI-assisted legal research, the growing adoption of generative AI by legal professionals renders similar procedural safeguards both timely and essential. Thus, comparative experience indicates that the most effective institutional response lies not in broad legislative reform but in embedding structured verification practices into the current procedural framework. This approach upholds existing legal principles while maintaining that judicial rulings are grounded in genuine and verifiable legal sources.
ETHINKING PROCEDURAL SAFEGUARDS: A VERIFICATION-CENTRED FRAMEWORK
The prior analysis shows that AI-created fake judicial authorities reveal a procedural flaw, not a gap in evidentiary doctrine. Therefore, the appropriate course is to enhance current procedural safeguards via systematic verification, rather than pursuing broad legislative reform. First, advocates must independently confirm each judicial authority found via AI-assisted research by cross-referencing it with an official court website, authorised law reports, or reputable legal databases before including it in pleadings or oral submissions. This obligation does not impose a new professional duty but reflects a modern application of the advocate’s established responsibilities of competence, diligence, and candour toward the court.[13]
Secondly, courts should promote reliance on authenticated legal reference systems when the legitimacy of cited cases is uncertain. Such a procedural safeguard would lower the chances that fabricated precedents sway judicial reasoning, without transferring the main duty of validation from counsel to justices. In the end, the Bar Council of India ought to provide professional guidance on the responsible use of generative AI in legal research. Following the American Bar Association’s approach, such guidance could set up practical verification procedures while emphasising that technology cannot substitute for independent professional judgment.[14]
Thus, a verification-centred framework enhances judicial integrity without disrupting existing evidentiary standards or necessitating broad statutory changes. By adding an integrated verification phase to legal research and litigation, current procedural protections can successfully adapt to the increasing use of artificial intelligence while maintaining the integrity of judicial authority.
CONCLUSION
Productive artificial intelligence has revolutionised legal research but also revealed a major procedural weakness in modern adjudication: the persistent belief that each cited legal precedent is authentic. As AI systems can now produce convincing yet completely fabricated precedents, relying on this assumption without independent verification is no longer tenable.
This article has contended that the issue posed by AI-generated sham precedents stems not from any inherent flaw in evidentiary law.Thus, the main contribution of this article is acknowledging that judicial integrity is most effectively safeguarded by implementing a structured confirmation process requirement for AI-assisted legal research. Enhancing procedural verification instead of broadening evidentiary rules provides a balanced and enduring solution that allows courts and legal practitioners to adopt technological advancements without jeopardising the integrity of judicial decision-making.
BIBLIOGRAPHY
Primary Sources
Statutes
Advocates Act, No. 25 of 1961, Acts of Parliament, 1961 (India).
Bharatiya Sakshya Adhiniyam, No. 47 of 2023, Acts of Parliament, 2023 (India).
Rules
Bar Council of India Rules, pt. VI, ch. II (Standards of Professional Conduct and Etiquette).
Cases
Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).
Gummadi Usha Rani v. Sure Mallikarjuna Rao, 2026 SCC OnLine SC 341 (India).
Professional Guidance
American Bar Association, Formal Op. 512 (2024).
Courts and Tribunals Judiciary, Guidance for Judicial Office Holders on Artificial Intelligence (Dec. 12, 2023).
[1] Gummadi Usha Rani v. Sure Mallikarjuna Rao, 2026 SCC OnLine SC 341 (India).
[2] Supreme Court Flags the Risks of Using AI for Drafting Petitions, DRISHTI JUDICIARY (June 2026), [https://www.drishtijudiciary.com/editorial/supreme-court-flags-the-risks-of-using-ai-for-drafting-petitions]
[3] Mata v. Avianca, Inc., 678 F. Supp. 3d 443, 448 (S.D.N.Y. 2023).
[4] The Bharatiya Sakshya Adhiniyam, 2023, § 56, No. 47, Acts of Parliament, 2023 (India).
[5] INDIA CONST. art. 141.
[6] Bar Council of India Rules, Part VI, Ch. II (Standards of Professional Conduct and Etiquette).
[7] Part VI, Ch. II, § I (Duty to the Court).
[8] American Bar Association, Formal Op. 512 (2024).
[9] The Contempt of Courts Act, 1971, § 2(c), No. 70, Acts of Parliament, 1971 (India).
[10] Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023).
[11] Am. Bar Ass’n, Formal Op. 512 (2024).
[12] Courts & Tribunals Judiciary, Guidance for Judicial Office Holders on the Use of Artificial Intelligence (Dec. 12, 2023); Law Society of England and Wales, Generative AI: The Essentials (2024).
[13] Advocates Act, No. 25 of 1961, § 49 (India); Bar Council of India Rules, pt. VI, ch. II.
[14] Am. Bar Ass’n, Formal Op. 512 (2024).
