Who Codes Justice? The Silent Use of Artificial Intelligence and the Challenge of Judicial Accountability in South-East Europe

In late 2023, UNESCO surveyed judicial operators around the world and found that 44 per cent were already using AI tools such as ChatGPT for work-related tasks. Only 9 per cent reported that their organisation had issued guidelines or provided training on such use. ¹  The distance between those two figures captures a defining governance challenge, which is that artificial intelligence has entered courtrooms and law offices well before institutions have decided how to receive it.

Across Europe, including South-East Europe (SEE), AI-based tools are gradually becoming part of everyday legal workflows, from administrative assistance and case management to legal research, document analysis, translation and various forms of decision-making support. ²  While I am not aware of any research-based figures for the region, my own communication with judges and attorneys suggests that such generative AI tools are already in use, oftentimes absent an approval, verification or other form of official endorsement in the courts. The pressing questions now concern how this technology is introduced, who controls its use, and whether legal institutions are prepared to safeguard the fundamental values upon which justice systems are built. ³

Recent scholarship on the digitalisation of justice shows that integrating AI into courts involves far more than technological modernisation. AI may contribute to efficiency, accessibility and improved court services, yet its deployment raises fundamental questions concerning fairness, privacy, cybersecurity, algorithmic bias and the preservation of human responsibility in judicial processes. ⁴  Considerable attention has accordingly been devoted to officially implemented AI systems in courts and public administration. ⁵  Important as this institutional perspective is, it leaves part of the picture in the shadows. A less visible, yet equally significant, dimension concerns what I have described in a recent paper as “silent AI use”, i.e., the individual adoption and use of AI-based tools by judges, lawyers and other legal professionals outside formally established institutional frameworks for their deployment, supervision and oversight. ⁶

What “Silent AI Use” Means

Silent AI use ranges from experimentation with general-purpose generative AI systems, which produce text or other content in response to a user’s prompt, to reliance on specialised legal tools selected by individual users. Its defining feature is the absence of institutional visibility, structured governance mechanisms and clear accountability frameworks. This says nothing about professional diligence because many professionals turn to these tools in good faith and out of professional necessity. Technological practices simply develop from the bottom up, through individual experimentation, before institutions have created appropriate rules and safeguards. The argument put forward here therefore concerns institutional capacity. Legal institutions cannot govern AI-supported legal work if they do not know where, how and which AI tools are being used.

The title question, “Who codes justice?”, builds on Lawrence Lessig’s famous insight that “code is law” meaning that digital environments shape human behaviour by enabling, restricting and directing choices, often without the visibility associated with traditional legal rules. ⁷  Technology companies undoubtedly code justice in a literal sense. They design the systems, select training methodologies and embed assumptions that influence how these tools operate. Yet as AI tools become increasingly accessible, a further question emerges: who decides which systems enter legal practice and under what conditions they are relied upon? Where courts and public authorities formally adopt AI systems, they can establish procedures for evaluation, procurement, security assessment and oversight. The widespread availability of general-purpose and specialised AI tools opens another pathway of individual selection and use by judges, attorneys, state attorneys, public notaries and other legal professionals. Through these choices, legal professionals themselves become actors who, in a less visible manner, code justice.

They neither create algorithms nor determine the technical functioning of AI systems. Their decisions about which tools to use, what information to entrust to them, and which outputs to incorporate into legal work nevertheless shape the technological environment in which legal reasoning takes place. When such choices occur without institutional visibility, common standards or accountability mechanisms, the process of coding justice risks shifting from a transparent institutional decision to a silent and fragmented practice. ⁸

Five Reasons Why Silent AI Use Matters

The first reason concerns confidentiality and privileged information. Legal professionals routinely handle sensitive information protected by professional secrecy, procedural confidentiality and, in many contexts, attorney-client privilege. Using external generative AI systems may involve transferring documents, factual descriptions or legal questions to third-party platforms whose data-processing practices remain opaque to the user. The CEPEJ Guidelines on the Use of Generative Artificial Intelligence for Courts, adopted in December 2025, warn that judicial professionals who use non-official generative AI services, whether free or commercial, risk inadvertent international data transfers, with the content potentially becoming available for model retraining and thereby escaping judicial administrative control. ⁹

The second reason revolves around personal data protection. AI systems may involve complex processing operations, including the collection, storage, analysis and potential reuse of personal data. Questions arise concerning controllers and processors, the legal basis for processing, transparency, data minimisation, security and data subject rights under the GDPR. The judicial context is particularly sensitive because it may concern vulnerable individuals, pending proceedings or special categories of personal data. The European Data Protection Board and national supervisory authorities have repeatedly emphasised that technological innovation must remain subject to the principles of necessity, proportionality and accountability embedded in EU data protection law. ¹⁰

The third reason relates to reliability of outputs. Generative AI systems do not “reason” in the legal sense. They generate responses based on patterns derived from training data and statistical prediction. This creates the well-known problem of hallucinations, where systems produce inaccurate, fabricated or misleading information while presenting it in a convincing manner, a problem that persists despite the advanced functions of newer AI tools. ¹¹  Courts have already sanctioned lawyers for filing submissions containing invented case law, most famously in the Mata v Avianca proceedings before a New York federal court ¹²  or issued regulatory ruling as in a case before the UK High Court. ¹³  Technological fluency cannot replace legal expertise and professional judgment, and AI-generated references, summaries or arguments require independent verification.

The fourth reason highlights the importance of judicial reasoning and explainability. Courts are required to reach decisions and to provide reasons that enable the parties and reviewing courts to understand their factual and legal basis. This obligation forms an essential component of the right to a fair trial under Article 6 of the European Convention on Human Rights, ¹⁴  a standard common to all Council of Europe member states in SEE. If AI tools contribute to legal analysis, research, drafting or the formulation of judicial reasoning, questions inevitably arise as to whether, and to what extent, that contribution should be disclosed, documented or capable of explanation. These concerns become particularly acute where general-purpose or proprietary legal AI tools operate as opaque “black-box” technologies whose internal processes users cannot meaningfully scrutinise. Responsibility for the final decision always rests with the judge. Meaningful accountability additionally requires that the process leading to that decision remain transparent and capable of review.

The fifth reason rests on accountability. The involvement of AI leaves untouched a fundamental principle of any legal system that legal responsibility remains with human decision-makers. Judges remain responsible for judgments, and lawyers remain responsible for advice and submissions. Meaningful accountability, however, presupposes that professionals adequately understand the tools they use and that institutions provide frameworks enabling oversight, guidance and evaluation. Without such structures, responsibility risks becoming diffuse because developers point to user discretion, users rely on technological authority, and institutions lack the information necessary to supervise either.

The South-East European Governance Paradox

Silent AI use is likely to be particularly relevant in SEE. Across the region, justice systems are undergoing digital transformation at markedly different speeds. Some jurisdictions have invested considerably in e-justice infrastructure, while others still face structural challenges relating to resources and technological capacity. Generative AI does not wait for institutional readiness. General-purpose AI tools are simultaneously available to judges, lawyers and legal advisers regardless of the level of digital maturity of their institutions.

This creates a distinctive governance paradox. SEE countries must modernise courts and legal services to improve efficiency, accessibility and public confidence in the administration of justice, while at the same time ensuring that technological innovation does not outpace the institutional safeguards necessary to preserve judicial independence, procedural fairness and fundamental rights. If governance frameworks lag behind adoption, the first widespread use of AI in legal practice may occur through individual choices well before any officially approved judicial system is deployed. Institutionally procured AI systems can, at least in principle, undergo legal, technical and ethical assessment before deployment. Silent AI use follows a different trajectory: readily available tools enter legal practice because they are efficient and immediately accessible, without corresponding safeguards for confidentiality, transparency or professional responsibility.

The regional diversity of SEE further complicates this picture. Croatia and Slovenia, as European Union Member States, are implementing the AI Act, which classifies AI systems intended to assist judicial authorities in researching and interpreting facts and law as high-risk, alongside existing obligations under the GDPR and broader EU digital legislation. ¹⁵  Other jurisdictions in the region are pursuing legal approximation through the accession process under different institutional capacities and timelines. AI tools, however, are available across borders irrespective of regulatory status. Legal professionals across jurisdictions may consequently use the same commercial AI systems within significantly different legal and institutional environments.

This asymmetry raises a broader question of technological inequality. Larger commercial law firms may possess the resources to procure specialised legal AI systems, develop internal governance policies and provide staff training, while smaller firms and sole practitioners may rely primarily on publicly available general-purpose tools. Similar disparities may emerge between courts with different levels of financial and technical capacity, producing unequal access to AI-assisted legal services. A further dimension concerns technological sovereignty. If legal institutions lack the resources or expertise to develop, procure or critically evaluate AI systems, they may become increasingly dependent on global technology providers whose models are trained, designed and updated outside European judicial traditions and often outside European regulatory oversight. At that point, silent AI use grows beyond an issue of professional practice and becomes a question of who ultimately shapes the technological infrastructure through which justice is administered.

From Silent AI Use to Visible Governance

The past two years have supplied SEE institutions with material to work with. The Council of Europe began addressing individual professional use as early as at the beginning of 2024, when the CEPEJ issued an information note on the use of generative AI by judicial professionals in a work-related context. ¹⁶  At the end of 2025, the CEPEJ adopted its Guidelines on the Use of Generative Artificial Intelligence for Courts, built on core principles that include the exclusive responsibility of courts for the exercise of judicial power, the non-binding character of generative AI outputs and the transparency of generative AI use in judicial activities and the drafting of decisions. ¹⁷  In the same month, UNESCO launched its Guidelines for the Use of AI Systems in Courts and Tribunals, the first global ethical and operational framework in this field, with recommendations addressed to judicial organisations and to individual judges alike. ¹⁸  For the legal profession, the CCBE has published guidance on lawyers’ use of generative AI. ¹⁹  These instruments share a premise that mirrors the analysis above: they speak directly to individual professional use, precisely because that is where adoption is happening first.

The task for every justice system goes beyond adopting AI, as it is necessary for the justice system to actually govern its adoption. This applies with full force to SEE, where several concrete steps follow. Judicial councils, court administrations and chambers should map the actual use of AI tools among their members, since governance begins with visibility. Ministries of justice and judicial authorities should issue interim guidance on confidentiality, data protection and verification duties before, and independently of, the procurement of official systems. Expectations concerning the disclosure of AI-assisted drafting deserve open professional discussion before case-by-case improvisation hardens into practice. Training, finally, should reach every level of the profession, from supreme courts to sole practitioners, if technological inequality is to be contained.

Officially implemented AI systems undergo at least a degree of regulatory scrutiny. The invisible and fragmented diffusion of AI through everyday legal practice escapes even that, and it is there that the greatest risk to public trust may lie. If justice is to remain transparent, accountable and centred on human decision-making, institutional governance must evolve at least as rapidly as the technologies that legal professionals are already beginning to use. Whether justice continues to be coded silently, or instead visibly and accountably, is a choice that institutions can still make, but the window for making it is narrowing rapidly.

* This blog has been written within the scope of the project “TECHNO-IUS: Legal and Ethical Implications of the Dehumanisation of the Justice and Public Administration through Legal Technology” (uniri-mzi-25-31), funded by European Union – NextGenerationEU. The views and opinions expressed are solely those of the author and do not necessarily reflect the official position of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them.

1. UNESCO, Guidelines for the Use of AI Systems in Courts and Tribunals (UNESCO, 2025), https://unesdoc.unesco.org/ark:/48223/pf0000396582 (accessed 8 July 2026). The survey results were first reported in UNESCO, Draft Guidelines for the Use of AI Systems in Courts and Tribunals (UNESCO, 2024).
2. European Commission for the Efficiency of Justice (CEPEJ), European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and Their Environment (Council of Europe, 2018).
3. See, e.g., Bullock, Justin B. et al. (eds), The Oxford Handbook of AI Governance (OUP, 2022); Valls-Prieto, Javier, Governance of AI in the Judicial System, in: de la Rosa, Esteban et al. (eds), Digitalization and Artificial Intelligence in Courts (OUP, 2025) 298-312.
4. Maas, Matthijs M., Architectures of Global AI Governance: From Technological Change to Human Choice (OUP, 2025).
5. See, e.g., Zalnieriute, Monika and Limante, Agne (eds), The Cambridge Handbook of AI and Technologies in Courts (CUP, 2026).
6. Kunda, Ivana, Implementacija sustava umjetne inteligencije u hrvatskom pravosuđu u svjetlu Akta Europske unije o umjetnoj inteligenciji, in: Bačić, Arsen (ed), Umjetna inteligencija i pravo (forthcoming in HAZU, 2026).
7. Lessig, Lawrence, Code and Other Laws of Cyberspace (Basic Books, 1999); Code: Version 2.0 (Basic Books, 2006).
8. Kunda, Ivana, op.cit. fn.6
9. European Commission for the Efficiency of Justice (CEPEJ), Guidelines on the Use of Generative Artificial Intelligence for Courts, CEPEJ(2025)18Final, adopted at the 45th plenary meeting of the CEPEJ (Strasbourg, 4-5 December 2025), https://rm.coe.int/cepej-2025-18final-en-draft-guidelines-on-the-use-of-generative-ai-for/48802a4ad1 (accessed 8 July 2026).
10. European Data Protection Board (EDPB), Opinion 28/2024 on certain data protection aspects related to the processing of personal data in the context of AI models (17 December 2024).
11. Liu, Patty, Stammbach, Dominik and Henderson, Peter, Who Checks the Citations? Benchmarking Legal Hallucination Detection (2026), https://doi.org/10.48550/arXiv.2606.21155 (accessed 8 July 2026).
12. Mata v Avianca, Inc., 678 F Supp 3d 443 (SDNY 2023).
13. Ayinde v Haringey [2025] EWHC 1383 (Admin).
14. Dymitruk, Maria, The Right to a Fair Trial in Automated Civil Proceedings, Masaryk University Journal of Law and Technology (12/2019) 27-44.
15. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), OJ L, 2024/1689, 12.7.2024, Annex III, point 8(a).
16. European Commission for the Efficiency of Justice (CEPEJ), Information Note on the Use of Generative Artificial Intelligence (AI) by Judicial Professionals in a Work-related Context (Council of Europe, 12 February 2024), https://www.coe.int/en/web/cepej/-/information-note-on-the-use-of-generative-artificial-intelligence-ai-by-judicial-professionals-in-a-work-related-context (accessed 8 July 2026).
17. European Commission for the Efficiency of Justice (CEPEJ), op.cit. fn.9
18. UNESCO, op.cit., fn.1
19. Council of Bars and Law Societies of Europe (CCBE), Guide on the Use of Generative Artificial Intelligence by Lawyers (CCBE, 2025), https://www.ccbe.eu/fileadmin/speciality_distribution/public/documents/IT_LAW/ITL_Guides_recommendations/EN_ITL_20251002_CCBE-guide-on-the-use-of-the-use-of-generative-AI-for-lawyers.pdf (accessed 8 July 2026).