Introduction and research methodology
A. Background and context
In India the judiciary continues to follow British procedures, not only in courtroom matters but also in matters such as dress codes, regardless of geographical conditions. In ancient India the administration of justice lay with the crowned king, who acted as judge. Over time, some regions of south India developed the Naattamai system, a form of rural justice under which the people of nearby villages appointed a reputed person as judge, known as the Naattamai, to settle disputes among them. This system operated without established courtrooms and was mostly held under a banyan or peepal tree before the entire public, reflecting the transparency with which disputes were decided. Some villages continued to follow this system for a few years even after independence. Later, owing to a decline of trust between people and to interpersonal disputes, many citizens turned to the formal judicial system inherited from the British, and disputes came to be resolved almost exclusively through the courts. As a consequence, the pendency of cases in the courts has risen over the years. The Indian judiciary and the government have established many courts to reduce this burden, and several initiatives have been taken to speed up trial proceedings. Among these, the introduction of AI in the judiciary appears inevitable. The concern of this article is that the risks associated with automated courtrooms staffed by AI judges that replace human beings must not be excluded from any assessment of the future of the justice system.
B. Research problem
The government continues to develop the Indian judiciary by establishing courts across the country, thereby reducing the burden on judges and addressing the grievances of litigants and victims. Artificial intelligence and other technologies have been introduced with the same objective. This article supports the use of AI in courtrooms as an aid to judges, but it opposes steps that would establish courtrooms in which disputes are decided by AI in place of human judges. Several questions arise. Can AI assess the situation of an accused person at the time an offence was committed? Can it interpret statutory provisions on its own where interpretation is required? Does the computer think for itself? Is its intelligence, in any true sense, in evidence at present? Can an AI be trained to pass orders or judgments? The public still seeks a speedy trial, but not an unfair one stripped of the human element.
C. Objectives and research questions
This work aims to scrutinise the effectiveness and the limits of AI in the judiciary. The applicability of emerging technologies is not in doubt; the concern is that they should be applied only once they have been well developed. Experimental deployments of technologies still under development can affect the life of an individual and may thereby implicate the right to live with dignity under Article 21 of the Constitution of India. UNESCO has published guidelines on the use of AI in courtrooms, but it does not advise that any human judge be replaced by AI. In that respect, a degree of regulatory framework has already been achieved.
The primary objectives are, first, to examine the outcomes of courtrooms supported by AI and other emerging technologies, and second, to explore, through the research literature, the AI-driven judgments delivered by computers that have replaced human judges.
The study is guided by the following research questions. Can artificial intelligence enhance judicial efficiency without compromising human discretion and judicial independence? Does the growing reliance on artificial intelligence in courts pose a threat to the human-centred nature of justice delivery? What safeguards are necessary to ensure that AI remains a tool for judges rather than a substitute for human adjudication?
D. Research methodology
This study employs a doctrinal and analytical research methodology. It entails a careful review of AI-enabled courts globally, of the policy framework developed by UNESCO, and of academic literature discussing automated courts that use AI. It also draws on certain case laws to strengthen its analysis. A comparative approach is used to evaluate the regulatory treatment of AI in courts. The study is not concerned with the technical questions of AI programs and design, nor with detailed regulatory regimes, but rather with the comparative study of how the human element should be preserved in particular circumstances.
Literature review
Professor Fredric I. Lederer, writing in 1997, described the courtroom in the following terms. The courtroom is a place of adjudication, but it is also an information hub. Outside information is assembled, sorted and brought into the courtroom for presentation. Once presented, various theories of interpretation are argued to the fact-finder, who then analyses the data according to prescribed rules, determined by the judge through research, analysis and interpretation, and determines a verdict and result. That result, often with collateral consequences, is then transmitted throughout the legal system as necessary. The courtroom is thus the centre of a complex system of information exchange and management. Because lawyers and judges deal continuously with data, high-technology courtrooms exist and virtual courtrooms are possible. Litigation is a dispute between two or more parties, resolved by a judge, jury or arbitrator following argument, usually by counsel. To prove facts in dispute, counsel present evidence: witnesses are called, their accounts heard and their demeanour studied, and physical evidence such as items, photographs, plans, schedules and video recordings is considered.1
This definition draws attention to the point that theories of interpretation are argued in the courtroom. Interpretation becomes necessary when there is no direct provision, or where an existing provision must be analysed closely. In such situations AI and other technologies have no capacity for independent reasoning: they can only return a result derived from what was fed in during programming. The mechanism can be understood through a simple combinatorial expression, the two-to-the-power-n minus one formula, written as 2^n − 1, where n is the number of variables. The formula gives the number of non-empty combinations that can be formed from n distinct elements, and it is straightforward to illustrate.
Consider three variables, x, y and z, so that the number of variables n equals three. Substituting n equals three into the formula gives two raised to the power three, minus one, that is (2 × 2 × 2) − 1, which equals 8 − 1, or 7. There are therefore seven possible non-empty combinations that can be formed, namely x, y, z, xy, xz, yz and xyz. In the same way, any answer to any question, in any arrangement of words, can be recognised by a system that maps the incoming sequence against combinations already held in its memory and returns the matching stored response. The speed of such matching is undeniably impressive. As the literature on automation observes, in many cases it remains necessary for a person to interact with an automated system in order to complete or add value to the creation of a document.2 This confirms that AI must be driven by human beings when decisions are to be made. The present state of development is not suited to decisions taken by the computer alone; it can only follow built-in, programmatic procedures.
The role of AI in courtrooms and its regulation
A. Efforts to address the concern through an enabling framework
While the advantages of introducing AI-based applications in the justice system are clear, there are also considerable risks associated with their use for automated decision-making and for predictive policing and predictive justice. The opacity of certain AI applications can create difficulties for the need to justify decisions, for the equality of arms between parties in judicial proceedings, and for other principles. Appropriate safeguards are needed to guarantee the protection of fundamental rights, including equal treatment and data protection, and to ensure the responsible, human-centric development and use of AI tools where their use is in principle appropriate. It is important that judgments be delivered by judges who fully understand the AI applications, and all the information taken into account by them, so that they can explain their decisions. It is therefore important that judges and prosecutors be trained in the use of AI applications.3
To secure better outcomes and to avoid backlogs, several bodies, including the CEPEJ and UNESCO, have drafted regulations and limits for the use of AI within courtrooms. The European Commission for the Efficiency of Justice of the Council of Europe adopted the European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and their Environment in December 2018.4 More recently, UNESCO developed its Guidelines for the Use of AI Systems in Courts and Tribunals in close collaboration with legal scholars and judicial training institutions around the world. These guidelines were formally launched at The Athens Roundtable on AI and the Rule of Law, held in London on 4 December 2025.5
B. What artificial intelligence actually is
John McCarthy first used the term artificial intelligence in 1955 in connection with the Dartmouth project, the aim of which was to have a machine simulate intelligence of its own.6 This prompts a question about what is now described as AI and whether the description is apt. On the basis of the combinatorial expression discussed above, it is clear that feeding a question to an AI application returns an answer already held in memory, identified by matching the word ranges of the query; the speed of that process is impressive, but it is not independent thought. On this view, it is worth asking how and when McCarthy’s proposal could truly be realised. The point can be tested by analogy with fields such as medicine, where the causes of the limited applicability of AI can be examined, and by asking how the results of AI-supported fields have been weighed in the public estimation. It is true that some countries, such as the United States and the United Kingdom, have applied AI in the judiciary and can point to the near-elimination of corruption; but the place of the human element in justice remains in question.
In the physical system, the judiciary is independent of the legislature and the executive and is empowered to exercise its discretion. It is difficult to see how such discretion could be entrusted to a computer. In particular, the capacity to weigh the situation in which an offence was committed is beyond an electronic machine. The interpretation of a provision such as the right of private defence can arise in many situations and remains the subject of debate when a court considers whether to convict an accused person.
The point is illustrated by Darshan Singh v. State of Punjab, decided by the Supreme Court of India. The Court observed that the right to protect one’s own person and property against unlawful aggression is inherent in a person, and that the duty of protecting the person and property of others is one that a person owes to the society of which he is a member, the preservation of which is both in his interest and his duty. As Jeremy Bentham said, it is a noble movement of the heart, that indignation which kindles at the sight of the feeble injured by the strong, a noble movement which makes us forget our own danger at the first cry of distress; it concerns the public safety that every honest person should consider himself the natural protector of every other.7 The Court cautioned that such protection must not be extended beyond the necessities of the case, lest it encourage a spirit of lawlessness and disorder, and that the right is accordingly restricted to offences against the human body and to those relating to aggression on property.8 This reasoning is among the clearest illustrations of the finely balanced decisions that arise in matters of private defence, in which the High Court and the Supreme Court weighed the accused person’s actions from different perspectives, taking account of human sympathy through the approach of a philosophical scholar. That exercise of judgment lies beyond the reach of a pattern-matching system.
C. AI as a legal assistant to judges
One of the most effective uses of AI in the judiciary today is simply helping courts to cope with the large number of cases they handle. In recent years, German courts have received an unprecedented volume of proceedings that overwhelmed the judiciary and resulted in delays. At the Stuttgart Higher Regional Court, judges working on these cases faced a backlog of more than 10,000 matters. The courts initially had no technology to cope with the volume; most work was done manually and was highly repetitive, and judges had to spend hours reading lengthy electronic pleading files that could run to hundreds of pages and differ only in a few case-specific features. The Ministry of Justice in Baden-Württemberg recommended using AI with natural language understanding and other capabilities to help categorise each case into the relevant case groups, and required a transparent, traceable system that protected data. IBM created an AI assistant named OLGA, which offered case categorisation, extracted metadata and could help bring cases to faster resolution. With OLGA, judges and clerks can sift through thousands of documents more quickly and use specific search criteria to locate relevant information; the system also contextualises the information that surfaces, preserves the case history and gives users a comprehensive view of the case and its origins. Relieved of highly repetitive tasks, judges can concentrate on complex issues, and the courts report that they anticipate the processing time of cases could potentially be reduced by more than 50 per cent.9
Having regard to these developments, it is important that AI be implemented and developed in courtrooms, but decision-making itself should not be transferred to any technology. AI can support judicial efficiency, but it must never compromise human discretion and judicial independence. It can streamline judicial workflows, yet it cannot supplant human reasoning. Consistent with this view, the Supreme Court of India’s draft Regulations for the Use of Artificial Intelligence in Courts, 2026 provide that AI is to remain strictly assistive, so as to preserve judicial independence.10
The threat to the human-centred nature of justice
A. Blind use may misguide the courts
A senior judge has warned that AI outputs should be considered, not relied upon, citing instances of fabricated case citations, described as hallucinations, and inconsistent AI responses during court proceedings. As he put it, one may consider AI, but one cannot depend on it entirely to make a judgment; AI tunes itself to the prompt it is given, and it lacks the emotions that allow a human to analyse a situation fully in legal cases. He referred to instances in which lawyers presented AI-generated material to courts, including one in which a fabricated Supreme Court judgment with fake citations was produced, an event described as a hallucination that led to the dismissal of the document and the initiation of administrative action. He emphasised that AI does not possess true cognition and merely detects patterns through data and algorithms, and he stressed the need to address the digital divide through practical measures, such as expanding e-service centres in remote areas to improve access to justice. The aim, he concluded, is not robot judges but cyborg judges, part human and part machine, who draw on the computational strength of AI while applying independent human reasoning; AI can assist, but it cannot replace judicial decision-making, and to bridge the digital divide technology must reach litigants directly.11
Beyond serving legal professionals, AI is expanding access to legal help for people navigating the legal system without a lawyer. Chatbots and virtual assistants can prepare legal materials and assist with governmental filings, which makes verification of AI outputs all the more critical. The use of AI carries both responsibilities and risks for legal professionals, who may be tempted to over-rely on AI output without adequate verification.12
B. Protecting the judicial system from bias by avoiding automated justice
The view set out by the CEPEJ in its Charter is that the application of AI in the field of justice can contribute to improved efficiency and quality, and must be implemented in a responsible manner that complies with the fundamental rights guaranteed, in particular, by the European Convention on Human Rights and the Council of Europe Convention on the Protection of Personal Data. For the CEPEJ, it is essential to ensure that AI remains a tool in the service of the general interest and that its use respects individual rights.13
C. AI cannot genuinely understand the matter before it
The Delhi High Court has expressly cautioned against blind reliance on artificial intelligence in judicial proceedings. In Christian Louboutin SAS v. The Shoe Boutique – Shutiq, Justice Prathiba M. Singh observed that AI-generated responses cannot form the basis for adjudicating legal or factual disputes, because the accuracy and reliability of such systems remain uncertain. The Court emphasised that AI may be used only as an auxiliary research tool and cannot replace human intelligence or the humane element inherent in judicial decision-making.14
The Delhi High Court’s guidance is that AI technologies may assist with legal research, document review, case management, access to databases and the filtering of long-pending matters, rather than with judicial decisions. As this article has shown through the combinatorial expression discussed above, AI responses are grounded in patterns and are not capable of the moral judgment, empathy and reasoning that are essential components of decision-making in the judiciary. Passing AI-generated orders or judgments risks producing inaccurate legal conclusions, altered citations, or outcomes that lose the genuine, distinctive substance of a dispute.
D. Global applications and challenges
In the Netherlands, a large central digitisation effort in the judiciary was cancelled for lack of progress, an evaluation committee citing, among other factors, legacy information systems as a technological challenge and difficulties in digitising analogue data and processes as a combined data, technological and organisational challenge. The Dutch judiciary has also pointed to a lack of funds as an economic challenge and to its dependence on the national budget; although the judiciary acts independently of the national government in theory, in practice its ability to act is constrained by the budget set by parliament, which presents a political challenge.15 Information about how an AI system functions is often withheld by manufacturers on grounds of operational secrecy, the protection of trade secrets, or concerns about the privacy of training data.16 The inability to explain how a model arrived at a particular decision can undermine the fairness of the process and may compromise a litigant’s right to a reasoned decision.17 The reasoning of an AI system cannot be held to the same standard as the reasoning offered by human judges.18
Conclusion
A. Key findings
This research seeks to strengthen the human element in judicial decisions that are increasingly constrained by artificial intelligence and other emerging technologies within so-called automated courts. Organisations such as UNESCO have issued guidelines on the use of AI in courtrooms. At present there are no parliamentary discussions in India directed at making a comprehensive regulatory framework for the applicability of AI, but juristic writing and the wider research literature support the view that litigation should be decided only by human judges rather than by automated mechanisms.
B. Addressing the doubts
It is important to recognise that digitalisation differs from automation. The digitalisation of courts can reduce the burden on human participants, especially judges and advocates. There is at present a heavy demand for cases to be represented by advocates across the world, and the introduction of video conferencing in the Indian judiciary has proved very helpful, allowing an advocate to represent a client before a court from wherever the advocate happens to be and to attend to multiple cases at the same time. The time that would otherwise be spent travelling can be used productively in case preparation. For judges, transcription time can be saved and redirected to the preparation of judgments, aiding the speedy disposal of cases. Automated courtrooms with self-deciding machines, however, cannot properly be countenanced, for they place the civic rights of the victim, the accused and the litigant at risk.
The applicability of AI in many fields, including any decision that must be taken with regard to the human element, will not be possible until the vision of artificial intelligence proposed by McCarthy is fulfilled. People should be aware of the pressure exerted on them to accept, as genuine intelligence, what is in fact an illusion. Industries keen to promote AI technologies cannot attain the level they claim, and so present the current state of the art as though it were complete, persuading the public through advertising and media that what exists is absolute AI. The result is that people become accustomed to treating machines as though they behaved exactly like human beings. In such an incomplete situation, there is a real danger in surrendering the judiciary to AI courts.
C. Suggested measures for AI implementation
Judges spend a great deal of time on routine, repetitive work while drafting judgments, consuming many valuable hours from the perspective of a litigant. As noted above, the queue of litigants at the courts grows longer by the day, and if the courts continue at the same pace, the failure to dispose of pending cases may cause the public to lose trust in the judiciary. Even in partition suits, in which the High Courts and the Supreme Court of India have held that the parties stand on an equal footing, a case may take a minimum of five years to conclude rather than ending in an ex parte outcome.
AI should be neither applied exclusively nor rejected entirely in the judiciary. Rather, a regulatory framework should be capable of being modified as society and human understanding advance. Its use should remain under human supervision, or be fully explored and reviewed before adoption. Such frameworks can enable courtrooms to use AI technologies while retaining public trust and safeguarding judicial discretion, accountability and human values in the administration of justice.
D. Recommended safeguards
These developments show that, as AI and other emerging technologies are used to improve access to justice, accessibility may be mistaken for reliability. The introduction of AI technologies into judicial and legal services must therefore be accompanied by an easily adaptable regulatory framework that can respond to contemporary concerns. The cautious effort to correct errors should never be made at the expense of justice and fairness in judicial matters.
Although the CEPEJ maintains that AI tools should remain instruments in the service of the public interest, practical application can diverge entirely from that aspiration. When courts come to depend, perhaps unconsciously, on algorithmic systems for legal research and for support in deciding matters, there is a serious risk that judges may defer to technological outputs. AI must therefore remain subject to continuous human review, because a judgment produced with its assistance may be inadequate to the demands of humanity and individual circumstance. AI is limited by its training data, and its reliance on large historical databases can yield results that are erroneous, partial, misleading or outdated. There is thus a real risk that computationally assisted decisions may repeat the errors, prejudice and favouritism found in previously human-decided matters. AI also lacks the capacity to recognise and adjust to contemporary societal change, or to play an active role in advancing the human rights of marginalised groups.
*****
Footnotes
1. Fredric I. Lederer, The Courtroom as a Stop on the Information Superhighway, 71 Reform 4, 4-9 (1997).
2. Suraj Govindaraj, Integrating Artificial Intelligence in Court Processes: Challenges/Opportunities and Issues/Possibilities slide 40 (Nat’l Judicial Acad. 2022-23).
3. Council of the European Union, Council Conclusions on Access to Justice – Seizing the Opportunities of Digitalisation, para. 35, 2020 O.J. (C 342) 1.
4. Eur. Comm’n for the Efficiency of Justice (CEPEJ), European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and Their Environment 1 (2018).
5. UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals (Dec. 3, 2025), https://www.unesco.org/en/articles/guidelines-use-ai-systems-courts-and-tribunals (last visited June 13, 2026).
6. John McCarthy et al., A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence (Dartmouth Coll. 1955).
7. Jeremy Bentham, Principles of Penal Law, in 1 The Works of Jeremy Bentham 297, 300 (John Bowring ed., 1843).
8. Darshan Singh v. State of Punjab, (2010) 2 S.C.C. 333, 348-49 para. 29 (India).
9. IBM, Judicial Systems Are Turning to AI to Help Manage Vast Quantities of Data and Expedite Case Resolution (Feb. 4, 2025), https://www.ibm.com/case-studies/blog/judicial-systems-are-turning-to-ai-to-help-manage-its-vast-quantities-of-data-and-expedite-case-resolution (last visited June 13, 2026).
10. Supreme Court of India, Draft Regulations for Use of Artificial Intelligence (AI) in Courts, 2026 (June 3, 2026) reg. 4.
11. Judges Caution Against Artificial Intelligence in Courts, Flag ‘Hallucinated’ Citations, The Times of India (Nov. 6, 2025), https://timesofindia.indiatimes.com/city/bengaluru/judges-caution-against-artificial-intelligence-in-courts-flag-hallucinated-citations/articleshow/125662932.cms.
12. TRI/NCSC AI Policy Consortium for Law & Courts, A Legal Practitioner’s Guide to AI & Hallucinations, Nat’l Ctr. for State Cts. (Feb. 16, 2026), https://www.ncsc.org/resources-courts/legal-practitioners-guide-ai-hallucinations.
13. Eur. Comm’n for the Efficiency of Justice (CEPEJ), European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and Their Environment 4-8 (2018).
14. Christian Louboutin SAS v. The Shoe Boutique – Shutiq, CS (COMM) 583/2023, paras. 21-24 (Delhi High Ct. Aug. 22, 2023) (India).
15. Lisa Julia Di Natale & Antonio Cordella, Digitising the Judicial Sector: A Case Study of the Dutch KEI Programme, in Electronic Participation 139, 139-53 (2023).
16. Frank Pasquale, The Black Box Society: The Secret Algorithms That Control Money and Information 3-18 (2015).
17. Danielle Keats Citron, Technological Due Process, 85 Wash. U. L. Rev. 1249, 1258-65 (2008).
18. Mireille Hildebrandt, Law as Computation in the Era of Artificial Legal Intelligence: Speaking Law to the Power of Statistics, 68 U. Toronto L.J. 12, 29-34 (2018).