Artificial Intelligence in the Indian Judiciary: Legal Accountability, Judicial Integrity, and the Emerging Regulatory Framework

I. Introduction

Artificial intelligence has permeated virtually every sphere of professional activity, and the legal profession is no exception. From automated legal research platforms to case management systems and predictive analytics tools, AI technologies are being deployed with increasing frequency across courtrooms, law chambers, and judicial offices throughout India. While these technologies carry considerable promise for improving the efficiency of a judiciary burdened by over 5.29 crore pending cases, they simultaneously introduce risks that strike at the foundational principles of judicial integrity, procedural fairness, and the rule of law.

The year 2026 has witnessed a watershed moment in this evolving landscape. The Supreme Court of India, in exercise of its suo motu jurisdiction, has taken direct cognizance of a deeply troubling phenomenon: the citation of fictitious case laws generated by AI tools in court proceedings. The Court's response, characterising such conduct as professional misconduct warranting legal consequences rather than a mere error in reasoning, signals a decisive shift in the judiciary's posture toward AI governance.

This article provides a comprehensive examination of these developments for the benefit of legal practitioners, judicial officers, and law students who must navigate this rapidly changing terrain with both competence and caution.

II. The Scale of AI Adoption in Indian Courts

2.1 Institutional Deployment of AI Tools

The Supreme Court of India and various High Courts have been active participants in the AI adoption movement, largely driven by the technological imperatives of the e-Courts Mission Mode Project. Several AI-powered tools have been institutionally deployed:

SUPACE (Supreme Court Portal for Assistance in Court Efficiency) is an AI-assisted research tool developed by the Supreme Court of India designed to process case files and surface relevant legal materials to assist judges in research. It is important to note that SUPACE is explicitly designed as an assistive instrument, and not as a decision-making mechanism. The tool remains in an experimental phase, pending further GPU infrastructure development.

SUVAS (Supreme Court Vidhik Anuvaad Software) has accomplished the translation of over 36,271 judgments into regional languages, significantly advancing the constitutional mandate of access to justice across linguistic communities.

TERES (Transcription and Recording of Evidence System) provides real-time transcription services during Constitution Bench hearings, reducing the clerical burden on court staff and improving the accuracy of official records.

In collaboration with the Indian Institute of Technology, Madras, the Supreme Court has also developed AI and machine learning tools integrated with the electronic filing system for defect detection in filings, and is presently testing prototypes for automated metadata extraction from judicial documents.

2.2 The e-Courts Phase III Investment

The Government of India has demonstrated institutional commitment to technology-driven judicial reform by allocating Rs. 7,210 crore toward the implementation of e-Courts Phase III. Of this allocation, Rs. 53.57 crore has been specifically earmarked for future technological advancement, including AI and blockchain applications across High Courts, with a projected implementation timeline extending to 2027. This investment represents the most substantial public sector commitment to judicial technology in India's constitutional history.

2.3 The Structural Context: Pendency Crisis

Any discussion of AI in Indian courts must be situated within the context of the pendency crisis that has long afflicted the Indian judiciary. As of mid-2025, over 5.29 crore cases remained pending across courts at all levels, including approximately 87,000 matters before the Supreme Court alone. This staggering backlog creates institutional pressure to adopt technological solutions, which in turn accelerates the deployment of AI tools at a pace that may outstrip the development of adequate regulatory safeguards.

III. The Crisis of AI-Generated Fictitious Citations

3.1 The Andhra Pradesh Trial Court Incident

The immediate catalyst for the Supreme Court's intervention was the discovery that a trial court in Andhra Pradesh had passed a judicial order relying upon four case laws that were entirely fictitious. These cases, which included purported citations such as Subramani v. M. Natarajan (2013) and Chidambaram Pillai v. SAL Ramasamy (1971), were not traceable in any recognised judicial database. Investigation revealed that these citations had been generated by an AI research tool and incorporated into submissions without any independent verification by the concerned legal professional.

This phenomenon, widely referred to in technical literature as "AI hallucination," describes the tendency of large language models to generate plausible-sounding but factually incorrect outputs, including citations to cases, statutes, and academic authorities that do not exist.

3.2 The "Mercy vs. Mankind" Episode and Broader Patterns

The Andhra Pradesh matter is not an isolated incident. A bench of the Supreme Court comprising the Chief Justice and two other Justices has flagged a disturbing pattern of AI-generated fictitious citations appearing across multiple proceedings. In one instance, a petitioner invoked a non-existent case titled Mercy vs. Mankind as a binding judicial authority. In another set of proceedings before Justice Dipankar Datta, not one but a series of fabricated judgments were cited in succession.

Perhaps more alarming than the citation of wholly fictitious cases is the practice of citing genuine Supreme Court judgments but attributing to them observations and holdings that those judgments never contained. This practice imposes a significant and unjustifiable burden upon judicial officers, who must undertake independent verification of every cited authority before addressing the substantive merits of a matter. The time and judicial resources diverted to such verification constitute a concrete harm to the administration of justice.

3.3 The Technical Explanation: Why AI Hallucinations Occur

For the benefit of legal practitioners and law students who may be unfamiliar with the technical underpinnings of this phenomenon, a brief explanation is warranted. Large language models, which form the foundation of most AI legal research tools, are trained on vast datasets of text. These models generate output by predicting statistically probable sequences of words based on patterns in their training data. They do not "know" the law in any meaningful sense; rather, they generate text that resembles legal writing, including plausible-sounding citations, case names, and judicial observations.

When a model lacks reliable training data on a particular legal question, it does not report uncertainty in a manner that a trained lawyer would recognise as such. Instead, it may generate a confident, well-formatted response containing fabricated citations. The output is syntactically coherent and superficially persuasive, making it particularly dangerous when used by practitioners who rely upon it without independent verification.

IV. The Supreme Court's Response: A Doctrinal and Regulatory Analysis

4.1 The Suo Motu Proceedings: Misconduct, Not Mere Error

On February 27, 2026, a bench of the Supreme Court comprising Justice P.S. Narasimha and Justice Alok Aradhe took suo motu cognizance of the Andhra Pradesh trial court incident. The Court's framing of the issue is of profound doctrinal significance. By characterising the citation of AI-generated fictitious case laws as "misconduct" warranting "legal consequences," the Court has moved the issue from the domain of procedural irregularity into the domain of professional and judicial discipline.

This characterisation has several important implications:

First, it signals that reliance on unverified AI output is not a pardonable error of judgment but a breach of the professional duty of care owed by legal practitioners to the court and to their clients.

Second, it places the verification of legal authorities squarely within the non-delegable professional responsibilities of the advocate or judicial officer who presents or relies upon them.

Third, it opens the door to disciplinary proceedings before Bar Councils and, in the case of judicial officers, to proceedings under the relevant service rules and constitutional provisions governing judicial conduct.

4.2 Institutional Stakeholders Engaged

The Supreme Court has issued notices returnable on March 10, 2026, to the Attorney General of India, the Solicitor General of India, and the Bar Council of India, and has appointed Senior Advocate Shyam Divan as amicus curiae. The engagement of these institutional stakeholders reflects the Court's recognition that the response to this challenge must be systemic, involving both the executive and the regulatory bodies governing the legal profession.

The Bar Council of India, as the apex regulatory authority for the legal profession under the Advocates Act, 1961, is particularly well-positioned to issue binding guidance on permissible AI use in legal practice. The March 10 hearing is widely anticipated to produce directions that will shape this regulatory landscape for years to come.

4.3 The February 2026 Supreme Court AI Guidelines

Prior to the suo motu proceedings, the Supreme Court of India had already issued comprehensive guidelines in February 2026 governing the use of AI in judicial administration. These guidelines articulate a nuanced position: AI is recognised as having the potential to enhance efficiency and accessibility within the justice system, while its limitations as a substitute for genuine legal reasoning are explicitly acknowledged.

The guidelines permit AI deployment in the following contexts:

• Case listing and docket management
• Legal research assistance (subject to independent verification)
• Translation and transcription services
• Data analytics for the identification of pendency patterns and case management insights

The guidelines expressly provide that AI shall function exclusively as an assistive tool and shall not substitute for judicial reasoning, judicial discretion, or the exercise of any adjudicatory function. This demarcation between AI-assisted administration and AI-driven adjudication is a principled and legally defensible position consistent with the constitutional requirements of judicial independence and the right to a fair hearing.

V. Professional Responsibility and the Duty of Verification

5.1 The Advocate's Duties to the Court

The professional obligations of advocates in India are governed primarily by the Bar Council of India Rules framed under the Advocates Act, 1961. Rule 11 of Chapter II, Part VI of the BCI Rules imposes upon every advocate a duty not to knowingly make false statements of fact or law. Rule 15 requires advocates to conduct themselves with dignity and to maintain the honour and dignity of the legal profession.

While these rules predate the emergence of generative AI, their application to AI-generated content is straightforward. An advocate who submits AI-generated citations to a court without independent verification cannot claim ignorance of their accuracy as a defence. The duty to verify the accuracy of every legal proposition placed before a court is a fundamental incident of professional competence and ethical obligation.

5.2 The Standard of Verification Required

In light of the Supreme Court's recent pronouncements, the minimum standard of verification required before citing any case law may be articulated as follows:

Every citation must be independently verified against a recognised and authoritative database, such as the Supreme Court's own judgment portal, SCC Online, Manupatra, or any other database that draws from official sources. The verification must confirm not only the existence of the case but also the accuracy of the proposition for which it is cited. AI-generated summaries of judicial holdings must be compared against the actual text of the judgment.

The use of AI as a preliminary research tool is not prohibited. However, the advocate bears full professional responsibility for every authority cited in court, regardless of the method by which it was first identified.

5.3 Implications for Judicial Officers

The obligations incumbent upon judicial officers in this context are, if anything, more exacting than those applicable to advocates. A judicial officer who incorporates AI-generated content into a judicial order without verification potentially compromises the validity and enforceability of that order, in addition to the disciplinary consequences identified by the Supreme Court. Judicial officers must be particularly vigilant in ensuring that any AI tool used in the preparation of judicial orders functions only as an aid to research and not as a source of independent legal authority.

VI. Towards a Comprehensive Regulatory Framework

6.1 The Regulatory Gap

Notwithstanding the Supreme Court's guidelines and suo motu intervention, India currently lacks a comprehensive statutory or regulatory framework governing the use of AI in legal proceedings. The Digital Personal Data Protection Act, 2023 addresses certain aspects of data governance but does not speak to the evidentiary, procedural, or professional responsibility dimensions of AI use in courts.

The draft National Data Governance Framework Policy and various sectoral guidelines issued by different regulators have not been harmonised into a coherent framework applicable to the legal and judicial domain. This regulatory lacuna creates uncertainty for practitioners, judicial officers, and technology developers alike.

6.2 Comparative Perspectives

India's regulatory challenges in this domain are shared by jurisdictions across the world. The Florida courts' mandatory disclosure requirements, the New York Unified Court System's comprehensive annual AI report, and California's SB 574 restricting AI use by attorneys all reflect a global convergence toward mandatory transparency and verification obligations. India has an opportunity to draw upon these international experiences in designing a framework appropriate to its constitutional structure and judicial culture.

6.3 Elements of a Prospective Framework

A principled regulatory framework for AI use in Indian legal proceedings should, at a minimum, address the following elements:

Mandatory Disclosure: Advocates and parties should be required to disclose to the court whenever AI tools have been used in the preparation of submissions, pleadings, or research materials. This disclosure obligation should be analogous to existing obligations of candour and would enable judicial officers to apply appropriate scrutiny.

Verification Certification: Advocates should be required to certify that all legal authorities cited in court submissions have been independently verified against authoritative primary sources, irrespective of whether AI tools were used in their identification.

Judicial Training: Systematic training programmes for judicial officers at all levels on the capabilities and limitations of AI tools are essential. UNESCO's recent survey finding that only 9% of judicial operators globally have received any AI-related training underscores the urgency of this need.

Bar Council Guidance: The Bar Council of India should issue binding guidance on the permissible use of AI in legal practice, incorporating minimum standards of competence, verification, and disclosure.

Technology Standards: Standards should be established for AI tools used in legal research and case management, including requirements for transparency, auditability, and source verification capabilities.

VII. Implications for Law Students and Emerging Practitioners

For law students and those at the commencement of their professional careers, the developments surveyed in this article carry specific pedagogical and practical implications.

The emergence of AI as a research tool does not diminish the importance of foundational legal skills. The ability to read and analyse primary sources, to trace the genealogy of a legal proposition through successive judicial decisions, and to evaluate the reliability of a legal authority are skills that remain indispensable precisely because AI tools cannot perform these functions with the requisite accuracy and reliability. Law schools must ensure that legal research and writing curricula equip students with the critical faculties necessary to use AI as a tool rather than as a substitute for legal reasoning.

Furthermore, law students must internalise from the outset of their professional formation that the duty of candour to the court is absolute and non-delegable. No technological tool, however sophisticated, can relieve an advocate of personal responsibility for the accuracy of the materials placed before a court.

VIII. Conclusion

The Indian judiciary's engagement with artificial intelligence represents one of the most consequential developments in the contemporary administration of justice. The deployment of AI tools such as SUPACE, SUVAS, and TERES offers genuine promise for addressing the pendency crisis and improving access to justice. However, the concurrent rise of AI-generated fictitious citations, and the Supreme Court's unequivocal characterisation of such conduct as professional misconduct, serves as a sobering reminder that technological capacity and professional responsibility must advance in tandem.

The March 10, 2026 hearing before the Supreme Court is poised to produce institutional guidance that will define the contours of permissible AI use in Indian legal proceedings for the foreseeable future. Legal practitioners, judicial officers, and law students must engage with these developments not as passive observers but as active participants in the shaping of a regulatory culture that harnesses the benefits of AI while preserving the integrity that is the foundation of the rule of law.

The central principle that must animate this regulatory endeavour is clear: artificial intelligence may assist the administration of justice, but it cannot and must not replace the human judgment, professional responsibility, and constitutional accountability upon which the legitimacy of judicial power ultimately rests.

References and Further Reading

• Advocates Act, 1961 and Bar Council of India Rules
• Bar Council of India Rules, Chapter II, Part VI (Professional Standards)
• Supreme Court of India AI Guidelines, February 2026
• Supreme Court of India, Suo Motu Writ Petition arising from Andhra Pradesh Trial Court AI Citation Matter (March 2026)
• e-Courts Mission Mode Project Phase III, Department of Justice, Ministry of Law and Justice
• UNESCO Guidelines on the Use of AI by Judicial Systems (2025)
• Digital Personal Data Protection Act, 2023
• SUPACE and SUVAS Documentation, Supreme Court of India