Comparative insights and future directions of AI in the courts of the Baltic States

Abstract

In the Baltic States—Estonia, Latvia, and Lithuania—a commitment to modernization and economic growth has led to an emphasis on digitization and the integration of artificial intelligence (AI) within public sectors, including judicial systems. Estonia, known for its technological advancement, leads in digital justice within the EU, while Lithuania and Latvia are making substantial progress. The relatively small population and court systems in these countries present both opportunities and challenges for AI deployment. This paper provides a comprehensive and practical analysis of AI implementation in the courts of the Baltic States, presenting the regulatory frameworks, policy documents, and specific examples of AI tools currently used in the judicial systems of the three countries. It also examines the future prospects of AI integration in these legal systems. The paper draws on publicly available sources, legal instruments, academic literature, and expert interviews conducted in the three Baltic States.

Introduction

The Baltic States—Estonia, Latvia, and Lithuania—experienced a turbulent history in the 20th century, marked not only by the World Wars but also by subsequent Soviet occupation. Regaining their independence in the 1990s, and joining the EU in 2004, the three countries have worked hard to grow their economies and modernize their public and private sectors.¹ This commitment to progress and innovation has led to their interest in adopting modern technologies, setting digitization as one of the key priorities² and seeking to harness artificial intelligence (AI) for public functions, including its application within judicial systems.

With Estonia gaining a reputation as a technologically avant-garde jurisdiction,³ and Lithuania and Latvia trying to catch up, all three Baltic States made various efforts to bring technology to courts. According to the Council of Europe European Commission for the Efficiency of Justice (CEPEJ) research measuring the functioning of the judicial system, in all three Baltic States the ICT deployment (Estonia’s—7.86, Latvia’s—7.57, Lithuania’s—6.10) and usage (Estonia’s—7.6, Lithuania’s—5.4, Latvia’s—5.12) indexes, measuring digital access to justice, case management and decision support, is above the Council of Europe median (4.16). Similar tendencies can be seen in the EU Justice Scoreboard 2024, which, inter alias, measures the digitalization of justice systems.⁴ According to it, in the use of digital technology by courts and prosecution services, Estonia is the first in the EU, while Lithuania and Latvia share 11th and 12th places.

The relatively small population and court systems in the Baltic States bring both difficulties and advantages for AI deployment. On the one hand, the limited scale of these systems makes them more adaptable to innovative solutions and facilitates faster pilot testing and implementation of AI-driven tools. On the other hand, these systems face constraints such as a smaller volume of data for training AI systems, which could impact the accuracy and reliability of such technologies, or usage of less-widely spoken languages, complicating fast integration of language models.⁵ Additionally, in smaller societies, the societal impact of judicial decisions is often magnified, requiring AI systems to meet particularly high standards of transparency, accountability, and fairness to maintain public trust in the rule of law.

This paper aims to offer a comprehensive and practical analysis of AI implementation in the courts of the three Baltic States, while also discussing potential future developments in this field. The paper starts with an overviewing of the regulatory background and presents the legal and policy documents on AI use in the public sector across the three countries. It then examines the concrete use of AI tools in courts in Lithuania, Latvia, and Estonia, discussing various examples of how technologies and AI tools are being integrated into judicial systems. Finally, relying on publicly accessible information and interviews with professionals, the authors reflect on the prospects for the future of AI in the courts of the Baltic States. It should be noted that the topic of court modernization and digitalization in the Baltic States has been discussed in research papers only to a very limited extent. For instance, Gamito Cantero and Gentile (2023) provided a practical overview of the Estonian context⁶; the Lithuanian perspective on the digitalization of justice was examined by Vėbraitė and Strikaitė-Latušinskaja⁷, as well as by Kiršienė et al.⁸ The existing body of literature on this subject, however, is often confined to discussions of automation rather than actual AI implementation and does not accurately reflect the most recent developments due to the fast-paced nature of contemporary AI. The authors therefore also aim to contribute to the currently limited scholarship on AI use in the Baltic States, particularly in English-language research.

Methodology

As its starting point, this paper relies on publicly available information and legal instruments in English and national languages, as well as relevant academic literature. To enrich the analytical part of the paper and to generate additional insights, this paper also incorporates the perspectives of professionals interviewed during the scientific project ‘Artificial Intelligence in Courts: Challenges and Opportunities’ (TeismAI), funded by the Research Council of Lithuania and implemented between June 2023 and May 2025 by researchers at the Law Institute of the Lithuanian Centre for Social Sciences, including us. The methodology of interviews is presented below.

To reach the policymakers and experts in relevant public institutions, official written requests to the institutions were made, and in some cases, the experts were contacted directly. To select academics, several authors of papers relevant to the area were contacted. Snowball sampling (when respondents suggest other respondents) was also used.

To ensure data collection consistency, the interviews were based on uniform interview guidelines developed by the researchers of the project. The guidelines were developed using existing literature, policy documents, and publicly available information. The Interview Guidelines included questions on the current use of AI in courts, areas where AI could be useful, judicial values and human rights, data privacy and security, transparency and explainability, accountability and responsibility, public perception and acceptance, training and skill development, and future trends and adaptability. Based on the respondent’s background, the interviewing researcher decided which questions to emphasize and which to address less extensively. Most questions were formulated broadly to allow the professional to discuss particular topics and points that were not foreseen by the research team and that correspond to the reality of the field.

Researchers followed the ethical rules applicable to their institution, and the approval of the Committee on Compliance with Research Ethics of the Lithuanian Social Science Centre was received for conducting the interviews. Prior to the interviews, the researchers shared the interview guidelines and all information related to the informed consent with the participants. Then, research project participants confirmed their participation by signing informed consent physically or through electronic means.

Ten professionals from the three Baltic States were interviewed by the authors of the paper. They were working for higher-level courts, public institutions tasked with court administration, or academics. The interviews were conducted via Zoom from January 2024 until December 2024. They were audio-recorded and transcribed. Each interview was assigned a code and the identities of respondents are anonymized in this paper.

There are certain limitations related to the interviews that should be noted. Firstly, not all individuals identified by the researchers as the most relevant respondents agreed to participate in the study. Significant challenges were encountered in finding qualified interviewees, due to the limited availability of professionals specializing in AI in courts in the Baltic States, their demanding schedules, and, more broadly, a general reluctance to participate in research interviews. Consequently, the perspectives of some key figures in the field may not have been captured. Secondly, the researchers had better access to professionals in Lithuania and Latvia, which resulted in a higher number of interviews being conducted in these two countries.

Regulatory framework for AI in the Baltic States

In the Baltic States, the use of AI is governed by both EU and national regulations. As EU Member States, Lithuania, Latvia, and Estonia adhere to a common EU legal framework, which includes key regulations such as the EU AI Act⁹ and the General Data Protection Regulation (GDPR)¹⁰. In addition to these EU instruments, each of the Baltic countries has enacted national regulations that complement and help implement the EU rules, addressing AI use in specific contexts.

In December 2018, the European Commission released a “Coordinated Plan on Artificial Intelligence”¹¹ (updated in 2021¹²), which, inter alia, encouraged Member States to develop national strategic visions on AI by mid-2019. Being political documents, such national AI strategies allow the reconstruction of how governments and policymakers view their role regarding technology and public responsibility for AI-based technologies;¹³ they also act as official roadmaps for the adoption and development of AI across various sectors of society.

Lithuania was the first of the Baltic States to publish its national AI strategy (published in March 2019)¹⁴, followed by Estonia (July 2019)¹⁵ and Latvia (February 2020)^16. In many aspects, the strategies of the three countries are similar, in particular as regards such areas as fostering AI adoption in public and private sectors, building AI-related skills and competencies¹⁷, and infrastructure development to support AI advancements.¹⁸ The AI strategies are implemented in specific action plans and other instruments.¹⁹ For instance, in 2022, in Lithuania, an ‘Action Plan for the Development of Artificial Intelligence Technologies in Lithuania in 2023–2026’²⁰ was prepared, seeking to ensure consistency in further work towards AI integration. Similarly, in 2024, Estonia published its ‘Action Plan for Artificial Intelligence 2024–2026’.²¹

Looking specifically at the judicial branch, AI implementation ambitions are evident in official strategy documents of courts within the Baltics. In Lithuania, the ‘Vision for the Development of Lithuanian Courts for 2023–2033’²² identifies several key areas of focus for the future years. The expansion of innovation in the courts, such as the creation and implementation of new technological solutions, including those using AI, is outlined as a specific action to be undertaken within the area of improving the quality of court processes and services. In Latvia, the priority budget measures of the Supreme Court for 2025–2027 include, inter alia, the introduction of AI in the provision of Supreme Court functions.²³

Apart from national instruments confirming the general openness to innovation in the courts, there are few specific instruments or guidelines addressed to courts and courts’ staff available, especially as regards possible generative AI use by judiciary.²⁴ However, courts would arguably benefit from soft-law instruments or internal policies, that can be described as the new modus operandi of modern technology governance.²⁵ This approach is already being followed by several states in the USA, Canada, New Zealand, and beyond, with courts adopting tailored rules concerning the use of Generative AI (GenAI) by judges and court staff.²⁶ Currently, no such instruments were found in the Baltic States.

Current applications of AI in Baltic courts

Case handling technologies

All three Baltic States employ technological solutions to manage case handling in courts. Active since 2004, LITEKO is the major judicial information system of Lithuania. This state information system is used for the efficient processing of procedural documents as well as relevant information and is the main tool for handling, accessing, and analysing cases. LITEKO has several functionalities and consists of the following modules: (i) the module of case registration and record; (ii) the module of public announcement of court decisions on the internet; (iii) the module of the generation of court statistics of and its placement on the internet; (iv) the module of exchange of case information among institutions; (v) the module of search of similar cases and information in courts; (vi) the module of court document templates; (vii) the module of preparation of court schedules; and (viii) the module of distribution of cases to judges.²⁷

Initially created to digitalize court documents, over the years LITEKO successfully adopted automation and has now reportedly introduced various AI elements into the recently modernized version of the system (major updating of LITEKO was implemented in 2023-2024). For instance, algorithm-based solutions in the distribution of cases module have been helping to ensure transparency through random case allocation for more than a decade. Recently it was enriched with an additional formula for calculating the judge’s choice factor and changed to ‘the case complexity scores’, which are important both for the allocation of cases and for determining the workload of judges set for cases.²⁸ In practice, however, this function is still viewed with some scepticism: “This has, of course, improved a lot over time, but it is still a major issue. The workload of judges is still uneven. That allocation of cases using AI, it falls far short of what is expected.” (Judge 3). Another respondent recognized the gradual enhancement of the case management system but noted that the updated LITEKO system would not lead to significant advancements in the judicial use of AI, whether in case preparation, legal research and analysis, or decision-making (Judge 2).

Estonian Court Information System (KIS2) is a similar tool to Lithuanian LITEKO. It handles everything from case registration and assignment to recordkeeping and online judgment publication. All information about the case is registered in the system, including the registration date, the classification of the case, the name of the judge or judges dealing with the case, the status of the case, all documents, hearings, and participants. KIS2 is also connected to the financial information system so that the payment of court fees and other court costs as well as payments made to the experts are all managed through KIS2.²⁹ The KIS2 system also enables automatic document generation. All court summonses are automatically produced using a centralized template, with clerks able to add warnings, explanations, and unstandardized text as needed. Clerks can also create personal or shared templates for generating other documents, such as basic court orders, side letters, or preambles of court decisions.³⁰ Moreover, KIS2 is linked to the e-File system which enables the simultaneous exchange of information between different parties. However, though overall advanced, the Estonian Court Information System is still criticized as technical flaws and recurrent glitches sometimes occur³¹ and the system consists of a burdensome amount of procedures that require mechanical document drafting and is far from perfect from a technological perspective.³²

In Latvia, court process-related matters are managed in the “E-lieta” e-case system (formerly known as manas.tiesas.lv). The portal is functionally similar to those of Lithuania and Estonia. E-lieta provides the opportunity for citizens to access several e-services and all case materials on the portal elieta.lv, such as rulings, minutes of meetings, and information about scheduled meetings. It is also possible to send submissions or claims, receive advice without court involvement, and access a collection of practical legal issues using an out-of-court guide.³³ The system is also equipped with a procedural fee calculator, up-to-date court calendars, and other accessible services.³⁴ Moreover, E-lieta includes virtual assistant “Justs” (see Section AI tools for communication with the public), and a tool “Robot” which ensures automated workflow for submitted e-forms.

Anonymization of court rulings in the public case law database

Court rulings anonymization (pseudonymization) tools were among the earliest automation technologies adopted by courts in the Baltic States. These tools were designed to ensure compliance with data protection regulations by automatically identifying and redacting personal data and sensitive information from court decisions before publication.

Digital solutions for court ruling anonymization have been available for nearly 20 years in a module of the Lithuanian judicial information system LITEKO. Considered to be one of the first automatization tools in the national judicial system, the system automatically converts specified names and titles into relevant symbols before publishing official documents. The results, however, must be thoroughly reviewed by a human. As noted by one respondent in the interview, “We call it semi-automatic because it is far from being perfect” (Judge 3). Although having been utilized for a while, there is little evidence that would suggest significant improvements in the foreseeable future.

A similar anonymization software that removes participants’ personal data from court judgements is used in Estonia.³⁵ In the initial stages of deployment, the program mistook names, personal identification numbers, and addresses, leading to the necessity of additional review by court staff. The tool struggled to recognize contextual hints and references to identifiable individuals. However, the error rate decreased as the system learned (Government Official 1). A text analytics tool created by Texta OÜ in Estonia is also reportedly being used by several government institutions to optimize work processes and streamline routine activities. For instance, in collaboration with the Centre of Registers and Information Systems, the Ministry of Justice used the system to remove personal data from nearly 80,000 court decisions involving outdated court sentences and republished the decisions in the Court Information System³⁶. Latvia also reports that by the end of 2019, 250,000 Latvian court decisions were available to the public after being anonymized using algorithms and selectively reviewed by the staff of the Court Administration.³⁷

Automated speech recognition systems for transcribing court hearings

Court hearings have traditionally relied on manual transcription, a labor-intensive process. Recent advancements in speech recognition and machine learning—a game changer for courts³⁸—now enable reliable automated transcripts, significantly reducing the time and effort required for preparing court documents. Unsurprisingly, all Baltic States are keen to introduce such tools in their courts.

Estonia started developing speech recognition technologies comparatively early and currently is applying them from their everyday utility within government departments (from chatbots to transcription services) to areas of emergency response.³⁹ In Estonian courts, the speech-recognition software Salme,⁴⁰ powered by natural-language technology, assists in recording court hearings by simultaneously generating a transcript alongside the audio recording of the session.⁴¹ During a court session, Salme generates a transcript of the proceedings with only a few seconds’ delay.⁴² While the system operates almost entirely automatically, some human input is required beforehand. This involves identifying the attendees prior to the session, enabling Salme to recognize individual voices during the hearing.⁴³ For this purpose, the operator manually adds the attendees to the court session and then assigns the microphones to the respective users.⁴⁴

The Latvian TILDE is one of the leading language technology companies in Europe. Launched in 1991, it develops neural machine translation systems, speech technologies, and custom chatbots. TILDE has been closely collaborating with Lithuanian and Estonian entities – for instance, during the 2024 presidential elections in Lithuania, Central Electoral Commission of the Republic of Lithuania partnered with the company and deployed an AI-driven chatbot to assist with incoming inquiries.⁴⁵ Moreover, the previously discussed Estonian speech-recognition software Salme has been developed in partnership with TILDE as well.⁴⁶ However, despite its success in Estonia, there is currently no official information about whether this tool or other similar language technologies are being employed in Latvian courts. An interview participant (Judge 1) noted that this may be due to the lack of vocal training data in Latvian and mentioned a nationwide initiative called “Balsutalka”⁴⁷, aimed at creating a digital dataset of spoken language. It entails people recording themselves pronouncing words or sentences and then submitting the audio files to the website. Authors state that the goal of this initiative was not only to enlarge the datasets but to make them more diverse in terms of speakers, accents, text genres and styles, intonations, grammar, and lexicon as well.⁴⁸

For several years, the Lithuanian Court Administration has been also planning to implement voice-to-text tools for use in courts. As in other Baltic States, developing voice-to-text tools for Lithuanian courts is challenging due to the language’s limited use and complex judicial vocabulary. Pilot projects conducted in several Lithuanian courts indicated the need for further refinement of the tool to improve its adaptability for transcribing proceedings in the Lithuanian language. However, similar AI systems are already used in neighbouring countries (e.g. Estonia and Poland), making implementation likely in the near future.

AI tools for communication with the public

AI tools have the potential to facilitate efficient communication between courts and the public. Through the use of chatbots and virtual assistants, the websites of courts can provide real-time responses to frequently asked questions, such as court procedures, case statuses, or filing requirements. Moreover, AI can be employed for preparing general court information for the public, such as press releases. The efforts to deploy such tools are visible in all three Baltic States.

In 2024, Vilnius University (Lithuania), in collaboration with the Supreme Court of Lithuania, launched a pilot project called TeDIA to explore AI-assisted automation of press release preparation for court decisions.⁴⁹ The project emphasized enhancing public trust in the judiciary by creating concise, clear, citizen-focused press releases.

TeDIA, a chatbot based on ChatGPT, was trained using detailed, court-specific instructions and developed with input from journalists, linguists, cybersecurity experts, and court staff. The user uploads in TeDIA a Word file with an anonymized court ruling and a prompt asking the chatbot to draft a press release. Once generated, the AI-generated draft must be reviewed, and the tool includes a reminder stating, “The final text must be reviewed and edited by responsible court officials.” Pilot testing at the Supreme Court of Lithuania began in January 2025⁵⁰, with internal court documents outlining ethical and responsible use guidelines developed.

Estonia is currently in the final stages of developing Bürokratt⁵¹ – the interoperable network of governmental chatbots in the form of a single virtual assistant, allowing citizens to receive information and e-services from several public authorities all at once. Initiated by the Ministry of Economic Affairs and Communications of Estonia and developed by the Department of Machine Learning and Language Technology of the Information System Authority of Estonia, the innovative solution would also provide users with the possibility of performing several tasks directly through the system via text or voice—from renewing identification documents to filing a claim.

At the end of 2019, the Latvian Court Administration launched a virtual assistant named Justs (from the word justīcija, ‘justice’), that helps users navigate the E-case portal of Latvia.⁵²The user-friendly chatbot answers inquiries about available e-services and consequently reduces the general workload of court staff. Just quickly performs technical and routine work, that otherwise requires a lot of resources. In the context of Lithuania, a similar virtual assistant named Justis is reportedly being trained to be integrated within the national court information system LITEKO⁵³, however, it is not in use yet.

AI-related topics in the training of courts’ staff in the Baltics

Notably, alongside the gradual development of policy documents and implementation of AI tools, the three Baltic States have recently initiated training programs for judges on AI-related topics. These programs aim to familiarize judges with the AI fundamentals, its applications in the judicial system, as well as ethical and legal considerations surrounding its use. In Lithuania, for instance, for the year 2025, the topics ‘Artificial intelligence tools and their potential for use in court’, ‘Using Artificial Intelligence in the work of a judge’, and ‘The digital era: the legal framework and artificial intelligence’ were included in the training programme for newly appointed judges.⁵⁴ In addition, there are also one-time events dedicated to judges, aimed at presenting them with innovations⁵⁵, events organized by the European Judicial Training Network,⁵⁶ moreover, specific subjects are also offered by various international organizations and initiatives.⁵⁷

During the interviews, the majority of respondents from Latvia, Estonia, and Lithuania reported an overall increase in AI-related training within the legal field and highlighted the importance of such training. Respondents emphasized the value of internal learning initiatives. ‘We are learning from each other’, explained one judge, highlighting the importance of colleagues sharing expertise through in-house presentations (Judge 1). Recognizing the interdisciplinary nature of AI in law, another interviewee, an academic, described collaborative efforts with colleagues to deliver AI-focused lectures to lawyers (Academic 1). Similarly, some respondents noted collaborative efforts between various national institutions to organize conferences for both the judiciary and the broader legal community (Government Official 1).

However, the interviews revealed some obstacles faced by national judicial systems in the context of training provision and (or) reception. They are related to the voluntary nature of the training, limited accessibility in the regions and lack of specifically designed training for judges and other staff working in courts.

Interestingly, the interviewed participants pointed out that AI literacy is not yet considered a mandatory skill for court staff but is encouraged. As one judge explained, “It is an additional capability... AI training is not part of centralized training courses, but rather a matter of individual interest for certain people” (Judge 2). Participation in available training is voluntary (the above-noted training program applies only to newly appointed judges), and another judge highlighted their own proactive efforts to seek out training independently, stating, “I participate in conferences, about which I don’t necessarily learn from the court network, I find them myself, mostly from the university...” (Judge 3).

This voluntary nature of training and the absence of sufficient promotion of training events may explain why some court staff members choose not to participate. As one judge observed, “some secretary or assistant who is not interested in these topics, they are not willing to participate, they read the title [of the event], they usually don’t even use that artificial intelligence or file classification directly at work, none of that…” (Judge 3). It also needs to be admitted that high workloads may leave judges and court staff with little time or motivation to attend elective training sessions, even if they are accessible.

Another AI-training-linked challenge is the fact that geographical disparities in AI training access exist, with limited opportunities available in regions outside of major urban centers. As one academic highlighted, ‘There is a lot of information about AI in Vilnius [the capital of Lithuania]. Various training events, seminars, opportunities to participate, ask questions, and get informed. Perhaps the problem concerns the regions; it is the biggest problem. They are less likely to receive information about AI and the use of this technology. Therefore, this issue should be addressed primarily through regional policy’ (Academic 2). This reveals, that to bridge the digital divide and ensure equitable access to AI training for court staff in all regions, even such small states as the Baltic countries should consider organizing regional training events and utilizing technology to deliver training remotely.

The last challenge mentioned during the interviews was the still under-developed design of the training material and its general nature. As described by one professional, webinars and conferences resemble an information campaign reflecting current trends but could not be considered structured training (Court administration representative 1). The effectiveness of AI training initiatives can be hindered by the inherent diversity of tasks within the court system. Given the wide range of roles and responsibilities, training needs to be tailored to specific AI tools and functions relevant to each position. Training that is overly general (insufficiently tailored to roles) may lack practical application for many staff members; and the constant evolution of AI systems also requires regular review and updates of the training materials.

Expectations and fears as to the future development of AI in courts

The attitudes surrounding the future of AI in courts are diverse. While possible scenarios are thoroughly analysed by many legal theory scholars and range from utopian to catastrophic, the interviewed professionals provide a more pragmatic point of view and share observations based on their own experience in judicial matters.

First of all, it is worth noting that, in addition to the non-judicial use of AI (discussed in Section 4), professionals already envisage that AI could assist with certain judicial tasks, and are eagerly following the AI developments at the same time acknowledging possible risks. While no mention was made to the possibility of delegating judicial decision making to AI, engaging AI as a court clerk was seen as a viable option. In particular, when estimating the possible role of AI tools in the near future, one respondent compared them to those of regular court staff, starting that AI could serve as an additional law clerk (Judge 1). On the one hand, this could possibly solve the issue of staff shortage, since, as another interview participant pointed out, “the labour market is not full of eager secretaries”, and it has become challenging to recruit translators or court clerks (Government Official 1). On the other hand, this could also mean that the previous number of assisting personnel might face redundance issues. Considering that “almost everything besides actual decision-making could be delegated to or assisted with by AI in one way or another” (Government Official 1), “AI might eventually become something (or someone) lawyers compete with” (Academic 1).

Interestingly, such an approach expressed by several interviewed Baltic professionals corresponds to the suggestions of a prominent scholar prof Volokh⁵⁸, who argues that the development of artificial intelligence is gradual, and its application in courts should reflect this. The professor suggests that the progression of AI technologies should be viewed similarly to a career path, where AI is “promoted” according to its advancing capabilities (from a secretary or translator to a legal assistant or judicial clerk, and from a judicial clerk to a judge or arbitrator). As revealed in the interviews, having already employed AI as secretaries for transcribing court proceedings and administrators assigning cases to judges, professionals from the Baltic States envision that AI could take the next step and handle some of the tasks typically assigned to judicial clerks.

Several such ‘clerk-type’ tasks were suggested by interviewed professionals as suitable for delegation to AI, and positive expectations were mostly directed towards concrete task automatization and improvement. Among tools eagerly anticipated by courts’ staff are AI-driven solutions for preparing summaries of case facts, conducting analyses of case law, and identifying and interpreting potentially applicable legal acts, including those derived from international and EU law.⁵⁹ Most research participants agreed that using AI to handle repetitive, time-consuming, and tedious tasks, such as summarizing, would significantly enhance everyday work efficiency. As one participant noted, “We currently see the most potential in using AI for productivity improvements” (Judge 4). These tools, however, require more attention in preparation and risk analysis (including evaluating whether concrete tools would not fall under high-risk AI under the EU AI Act) as compared to those used for automation of non-judicial tasks since, for instance, selecting a case law to be followed might significantly influence the judge and its decision. Moreover, Baltic professionals expect that AI systems could help organize and categorize data systematically, ensuring that information is easier accessible and manageable.⁶⁰ As noted by one judge, managing cases consisting of large volumes of complex legal documents currently require significant human effort, but could potentially soon be organized by machines and would only take minutes to complete (Judge 4).

The interviews reveal that AI implementation could influence the future of courts in several ways. As is evident from most previous technology breakthroughs, what was once considered a novelty, tends to gradually become mainstream. One respondent expresses excitement and notes that contrary to sci-fi-related predictions, AI-powered tools will perhaps become just a regular, absolutely normal tool to use at work, as is with computers (Lawyer 1). Some participants even believe the potential transformative effect of AI in courts to be an overestimation of reality. While private law firms lead in AI applications, courts as state institutions will try to keep up but unfortunately remain constrained by bureaucratic processes, such as public procurement procedures or long-term projects. Having taken past experiences into consideration, one respondent states: “[AI implementation in courts] might happen, but I would not expect some special breakthrough, not anytime soon. No matter how pessimistic it sounds, I am just speaking from experience, having previously seen similar attempts…” (Judge 3). Some, however, expect a more significant change. As suggested by one respondent, even the process of drafting law will become strongly integrated with intuitive and context-aware AI systems. This would essentially turn law into machine-readable code (Academic 1).

When asked about future prospects, another professional brought to attention the risks generative AI raises the question of evidence evaluation and admissibility in courts and the need for the courts administration to react to that: “Is it generated text, images, videos, sounds, are they original? How should we verify that? We do not have the tools; we do not have the capabilities yet” (Academic 2). A potential solution to this issue would probably be the development and adoption of advanced forensic technologies (again, AI-based) capable of detecting AI-generated content, such as deepfakes or synthetic media, as well as establishing clear guidelines for assessing the authenticity and admissibility of digital evidence before the courts.

Finally, one of the most controversial aspects of AI in courts is widely debated in literature and concerns the core of justice—judicial decision-making.⁶¹ In this context, respondents refrain from speculating and emphasize that courts must draw the line at allowing AI systems, especially freely available ones, to replace the reasoning of the judge (Judge 4). Otherwise, as noted by many respondents, upholding judicial values, maintaining public trust in courts and several other aspects inherent to the nature of justice would be under threat.⁶²

Regardless of the methods and extent to which AI will truly establish itself within the judicial system, most respondents agreed that it is no longer possible to disregard the changes ahead. Even though courts are typically perceived as an ‘ancient and conservative system’ (Academic 1; Judge 1), the transition towards AI utilization in judicial settings appears to be inevitable. Its success does, however, entirely depend on the direction set by humans.

Concluding remarks

The current state of AI-based automation in the Baltic States’ judicial systems demonstrates that AI tools have so far primarily supported court staff by facilitating tasks such as anonymization, transcription, registration, and the issuance of standard documents, rather than directly affecting the work of judges. When AI is just entering courts’ work, such emphasis on non-judicial tasks is logical, as these areas present significant implementation potential and relieve courts of technical tasks with minimal controversy; they do not pose more substantial legal risks or ethical concerns. However, as the interviews revealed, professionals are already envisioning that in the near future, AI will extend into judicial decision-making and assist with tasks that are currently typically assigned to court clerks. Given the eagerness of the three Baltic States to advance gradually with technology, they are quite likely to be among the first in the EU to develop and deploy AI tools in partial automation of judicial decision-making.

With the increasing use of AI in courts, however, training court staff on AI usage deserves greater attention in the Baltic States (as in many other European countries), since the region does face challenges related to AI literacy. Judges in the Baltic States typically receive training only when entering the profession, specializing or adapting to legal changes, but this is insufficient in the face of advancing AI technologies. Mandatory introductory and continuous AI training is necessary to help judges and other court staff understand AI basics, its possibilities, and risks, to recognize potential AI use in the case materials, as well as to engage with societal and legal debates. Moreover, as AI continues to play a more prominent role in the legal field, there is an opportunity for the Baltic countries to follow suit, developing their own internal guidelines or soft-law frameworks to effectively govern the use of AI in their courts. This would not only enhance the adoption process but also ensure that the integration of AI aligns with the values of fairness, transparency, and justice.

Funding

This work was supported by the Research Council of Lithuania (LMTLT) under Grant ‘Artificial Intelligence in Courts: Challenges and Opportunities’ (TeismAI), agreement number S-MIP-23-73.

Footnotes