ABSTRACT

1 INTRODUCTION

This opinion article examines the anticipated impact of AI on judicial independence. It is premised on the belief that the judicial function is a core state responsibility and that, even with the proliferation of online alternative dispute resolution, the legitimacy and trust in state authority will always rest on judicial independence. The rise of AI and its use, whether openly acknowledged or not, by judges introduces new threats and questions for judicial independence. While AI can improve the administration of justice and ease litigation backlogs, its applications, ranging from assisting in judicial decision-making and case allocation to enabling virtual hearings or even substituting for judges, pose new challenges.

The pivotal question is whether individual judges should be permitted to use AI of their choice, or if the national or supranational justice system should provide a single, verified AI platform based on controlled, transparent datasets to ensure quality and trust. At the current stage of evolution, judges might, even with an official system available, opt for their own AI tools, which could result in significant equality problems, introduce biases due to the diversity of AI systems, and lead to errors influenced by the data used. Furthermore, using AI outside a secure, dedicated system could compromise data protection and the confidentiality of judicial deliberations. Nevertheless, establishing such secure systems requires considerable investment, which is challenging given the already strained resources of any justice system.

The use of AI also introduces additional challenges. Judicial reforms based on AI can reinforce or weaken judicial independence and could create avenues for undue influence over the judiciary. Examples of such influence include establishing liability regimes for judges using AI, linking judicial salaries to performance, or cutting judicial funding once AI systems are introduced. Control by the executive or legislative branches over the management of data spaces, the selection of training data, or the design of court algorithms could raise doubts about the judiciary's autonomy. The technical complexities of AI require specialist knowledge and the recruitment of professionals to translate legal rules into digital processes. There is also the question of whether judges themselves should be trained in AI or rely on technical experts, and how liability should be distributed between judges and AI technicians. A joint liability regime may be appropriate, but it must not undermine judicial independence.

2 ON JUDICIAL INDEPENDENCE: FOUNDATIONAL PRINCIPLES

The notion of judicial independence, much debated recently, rests on two basic assumptions: first, judges must be impartial and free from external pressures, whether from the media, litigants, or even other judges; second, independence ensures fair decisions based solely on evidence and law. These principles have been codified in several standard-setting global declarations,¹including the UN Basic Principles on the Independence of the Judiciary²and the ELI.³

Judicial independence is essential for democracy and the rule of law. Judges must not only be independent but also demonstrate independence to the public.⁴Their independence justifies the immunity they enjoy for actions taken in court. Judges are, though, evaluated periodically and can even be removed for misconduct. They are also liable to declare personal interests and avoid conflicts of interest or perceived bias.

3 HISTORICAL EVOLUTION OF JUDICIAL INDEPENDENCE

The roots of judicial independence trace back to the Act of Settlement in England and Wales in 1701⁵and later to the Enlightenment on the continent, notably to Montesquieu's “De l'esprit des lois” (1748),⁶which first articulated the separation of powers as a safeguard against despotism. Gradually, during the Nineteenth Century, this principle took root in European constitutions.

After World War II, judicial independence was codified at the international level through Article 6 of the European Convention on Human Rights (ECHR).⁷Within the European Union, the Court of Justice of the European Union (CJEU) elevated judicial independence to a constitutional principle, grounded in Article 19 of the Treaty on European Union (TEU)⁸and Article 47 of the Charter of Fundamental Rights.⁹The CJEU considers judicial independence essential for national courts to safeguard EU rights and to avoid violating the rule of law.

4 JUDICIAL INDEPENDENCE IN THE EUROPEAN LEGAL LANDSCAPE

The principle of judicial independence is a manifestation of the broader EU principle of effective judicial protection. Under this principle, Member States should offer effective legal remedies in the fields covered by EU law, including the guarantee of independent courts. Judicial independence, as reaffirmed by the CJEU in "Associação Sindical dos Juízes Portugueses" (C-64/16),¹⁰requires that courts be free from both political and external pressures, including, increasingly, technological dependencies. In "Commission v. Poland" (C-619/18),¹¹the Commission raised concerns about the recent reform empowering the Polish Minister for Justice to authorise judges to continue performing judicial duties beyond the retirement age. The CJEU found that these procedural rules failed to protect judges from potential direct or indirect influences on decision-making and thus that Poland was in breach of Articles 19(1) of the Treaty on European Union and 47 of the Charter of Fundamental Rights. In the ruling in case C-824/18,¹²the CJEU considered that judicial independence and impartiality presuppose rules, particularly regarding the composition of the body, the appointment, length of service, and the grounds for abstention, rejection, and dismissal of its members, to ensure public trust in the justice system.

The impact of artificial intelligence on judicial independence was raised in a recent request for a preliminary ruling lodged with the Court of Justice of the EU (Case C-159/25, Rowicz).¹³The case questions whether Poland’s system of random case allocation and judicial assignment, via an AI algorithm controlled by the executive, undermines judicial independence, impartiality, and fair trial rights under EU law. Advocate General Spielman delivered his Opinion on 19 March 2026¹⁴whereby he considered that Article 19(1) TEU, read in the light of Article 47 of the Charter of Fundamental Rights of the European Union, precludes national legislation allowing to withdraw cases allocated to a judge in order to reallocate them to a different formation, without laying down the criteria to be followed nor requiring a statement of the reasons, but that it does not preclude national provisions to organise random case allocation by an automated system, involving no human intervention, provided that that system has safeguards to prevent any arbitrary use and any undue interference.

5 THE AI ACT AND RISKS TO JUDICIAL INDEPENDENCE

The EU Artificial Intelligence Act (AI Act),¹⁵adopts a risk-based approach, classifying AI systems used in the justice sector as “high-risk.” This designation mandates strict safeguards to ensure human oversight and explainability. The underlying principle is that AI can assist but can never replace human decision-making in the judiciary. On top of that, the requirements outlined in the AI Act apply not only to systems that support judicial decision-making but also to tools used for case management and judicial administration.¹⁶Agentic AI is a clear direction that will influence the integration of AI into the judicial system, automating case handling beyond current case management and adding problems beyond the existing framework.

A critical aspect of the Act involves the roles of AI system providers and oversight entities. Whether these are commercial organisations or government branches overseeing the configuration or operation of an AI system, their involvement could significantly impact judicial independence.

Additionally, the AI Act places high-risk AI systems under the jurisdiction of various entities responsible for conformity assessment and governance, including the European Artificial Intelligence Board, national competent authorities, and market surveillance authorities. This raises the question: how will these entities interact with judicial administration? This interaction raises concerns regarding the potential effects on judicial independence.

Particularly noteworthy is Article 29 of the Act, which mandates that those using high-risk AI systems ensure that the individuals responsible for operating them have received adequate training.

Finally, this shift introduces a liability paradox. The AI Act emphasises the need for human oversight of high-risk systems, especially judicial AI. However, it raises an important question: how is liability allocated if a judge endorses a problematic AI recommendation? Would a shared liability regime be appropriate? Importantly, any liability framework for judges in AI-assisted courts should avoid exerting influence over their judicial functions. In practice, the threat of liability may pressure judges to rely on AI recommendations, thereby subtly coercing them and undermining their autonomy.

6 HUMAN RIGHTS IMPLICATIONS OF AI IN JUDICIAL SYSTEMS

The integration of AI in judicial decision-making raises significant human rights concerns, particularly regarding the right to a fair trial and the provision of reasoned judgments.¹⁷AI systems often lack transparency, making it difficult for judges to explain AI-generated decisions, thereby violating litigants' rights to understand the reasoning behind judgments. There is a risk that AI's reuse of historical data could reinforce preexisting biases, obstruct adaptation to more recent social changes, and potentially harm marginalised groups. The Council of Europe's CEPEJ European Ethical Charter on the Use of AI in Judicial Systems¹⁸articulates five guiding principles: respect for fundamental rights, non-discrimination, quality and security, transparency, and user control.

The European Court of Human Rights has underscored the need for legal frameworks that prevent executive interference and ensure transparent procedures for the appointment of judges. The context in which a judicial reform occurs may disclose an undue intrusion into the realm of judicial independence. Moreover, it reaffirmed the necessity of respecting the law in establishing tribunals, as it helps set out clear procedures for appointing judges. It thus prevents potential interference by the executive with the judicial power.

7 CONTEMPORARY ISSUES: THE UN REPORT ON AI IN JUDICIAL SYSTEMS

On 16 July 2025, the UN Human Rights Office published report A/80/169, "AI in judicial systems: promises and pitfalls",¹⁹authored by Special Rapporteur Margaret Satterthwaite. The report recognises the widespread adoption of AI in judicial systems worldwide and notes that many judges are already using AI tools, often in an ad hoc manner. The Rapporteur warns against "techno-solutionism", urging states and justice professionals to scrutinise the justice problems AI can help address, to assess the suitability of AI, and to consider alternative, less risky solutions. The report advises that AI should only be adopted after careful consideration of potential harms, strategies to mitigate those risks, and an evaluation of whether other solutions may be preferable.

The Rapporteur also emphasises the necessity of access to a human judge for the right to an independent and impartial tribunal, and to a human lawyer for the right to legal counsel of one's choosing. Therefore, the judiciary must be responsible for adopting any innovation that could affect judicial decision-making. To preserve judicial independence, judges must be trained to become proficient in digital and AI technologies and empowered to consult technologists, legal experts, and the public about the appropriate use of AI systems. Once decisions regarding AI adoption are made, essential information about the systems must be made publicly available.

The intersection of AI and judicial independence is thus a complex domain in which the benefits of technological assistance must be carefully balanced against risks to fundamental judicial values. As AI systems become more advanced, there is an increasing need for comprehensive ethical frameworks and robust regulatory oversight. Any integration of AI into judicial processes should be governed by transparent guidelines, ongoing evaluation, and a steadfast commitment to both the spirit and letter of judicial independence.

8 AI IN THE COURTROOM: EMERGING ISSUES

The implementation of AI in the courts raises additional issues for judicial independence.

Judicial reforms, often including cutting judicial funding and introducing performance measures, are likely to affect independence, meaning judges should be safeguarded from interventions or pressure by a state's executive or legislative branches, especially procedural rules that could influence their decisions, even indirectly.

Moreover, due to the technicalities of AI, implementing ICT solutions requires expert knowledge and, in practice, the recruitment of new professionals to translate rules and practices into computerised processes and datasets. Thus, a complex configuration is expected to oversee Court operations, comprising judges trained in AI, capable of assessing datasets and explaining the reasoning behind their conclusions, as well as supporting technicians responsible for operating AI in the courtroom, or a combination of both.

To train judges, a curriculum of the knowledge and skills required to address AI issues needs to be drafted. Since we are still in an uncoordinated stage of AI implementation, the answer is unclear. Under the aforementioned second scenario involving AI technicians, guarantees of their independence are crucial. Any technical expert working in this field should be free from any pressure or interference of the same intensity as that faced by a judge. A related question is the relationship between the technician and the judge, as well as the distribution of liability. At the same time, any liability regime for judges in AI-assisted courts should not entail any influence over judicial functions.

Ultimately, AI artefacts for the judiciary must be designed as assistive tools for judges, not as unsupervised adjudication instruments. The principle of judicial independence demands that the final decision must always be a human one. Nevertheless, even such a stance vis-à-vis AI remains simplistic, since the judicial function cannot be seen merely as confirming the reasoning of an AI system, but rather as a human intellectual-logical path that should also be explicated in the judgment delivered.

Judges must use AI to augment their judgment, not substitute it. They must critically evaluate AI-generated recommendations, much as they would assess the testimony of an expert witness, accepting or rejecting them based on reasoned analysis. At the same time, AI systems should be designed so as not to impose paths of reasoning. The latter should also allow judges' discretion to be part of the reasoning, grounded in human qualities of empathy and equity and taking into account the humane context and the commitment to individualised justice.

9 THE STRESS ON JUDICIAL INDEPENDENCE IN THE AI ERA

Judicial independence extends to judicial methodology. Judges enjoy considerable discretion in how they research precedents, weigh evidence, and arrive at conclusions. One could, on that basis, pretend that a judge should be free to choose which AI system to use, if any. However, AI introduces a new dimension to this independence, transforming discretion into risk.

Any judge independently employing an AI system is outsourcing elements of judicial reasoning to opaque technological systems whose inner workings, training data, and embedded biases remain hidden not only from litigants and the public but also from judges themselves.

This creates what we might call the "black box paradox": a contradiction in terms of judicial independence, a principle meant to ensure transparent accountability in a judge's reasoning, but that, despite all, permits the use of a system that advises the judge without the latter clearly understanding his own decision-making process.

10 BALANCING INDEPENDENCE AND EVALUATION

Finding a balance between judicial independence and evaluation processes is equally essential for effective justice. Judicial independence must remain the key orientation, since evaluations aim to improve judicial quality and identify training needs. The Consultative Council of European Judges (CCJE) in its Opinion No. 17 (2014)²⁰addresses the evaluation of judges' work, emphasising the balance between judicial independence and the quality of justice, and the importance of maintaining this balance to uphold the rule of law.

Judges have essential responsibilities that must be assessed to ensure the decisions they deliver meet the required standards. They must protect rights and freedoms equally and provide efficient dispute resolution. They are required to decide cases independently and on time, with clear reasoning. Judicial independence does not exempt judges from accountability; evaluations should aim to enhance the judiciary's level of competence. Nevertheless, the introduction of AI into the Courts' operations raises the question of whether judges' evaluations include assessing their capacity to use AI tools and further clarifying their degree of AI involvement.

11 THE CASE FOR INSTITUTIONAL CONTROL OF AI IN JUSTICE

AI technologies are increasingly utilised in judicial processes across various countries, promising to enhance efficiency and reduce costs.²¹Proponents argue that AI can deliver fairer, more consistent judgments compared to human judges, who may exhibit inherent biases and inconsistencies. AI tools can predict case outcomes and assist judges by analysing vast amounts of data, potentially transforming the judicial landscape by streamlining processes and improving access to justice.

Courts cannot avoid the AI issue. On the contrary, the trend should be to implement AI solutions to attract genuine benefits: enhanced legal research capabilities, pattern recognition across vast case databases, identification of relevant precedents that human researchers might miss, and administrative efficiencies that free judicial time for the human judgment that genuinely matters.

Nevertheless, the integration of artificial intelligence into judicial systems represents one of the most consequential transformations in modern jurisprudence. As courts worldwide experiment with AI tools, they face a fundamental question: should individual judges have the autonomy to adopt AI systems ad hoc, or must these technologies be provided and regulated exclusively by state judicial authorities?

The answer is unequivocally the latter. For as long as judicial independence remains a cornerstone of democratic society, the unregulated use of AI by individual judges threatens to undermine the very principles that independence is meant to protect: equality before the law, transparency, and the consistent application of legal standards.

Nevertheless, this cannot happen in disorder; it requires institutionalisation. State judicial authorities, supreme courts, judicial councils, and ministries of justice must assume exclusive control over AI deployment in courtrooms. This implies centralised procurement and validation. Collective judicial authorities, not individual judges, should evaluate, select, and procure AI systems. This allows for rigorous testing for bias, validation of accuracy, assessment of interpretability, and negotiation of transparency requirements with vendors.

Mandatory training and standardisation ensure all judges receive uniform training on the AI tools approved for use, ensuring consistent application and a shared understanding of their limitations. The tools themselves should be standardised across jurisdictions to prevent the inequality and incoherence that ad hoc adoption creates.

Transparency and accountability mechanisms require that institutionally deployed AI operate under clear rules that require disclosure when AI systems are used in case analysis, document the role they played in judicial reasoning, and provide opportunities for litigants to challenge AI-assisted decisions. Regular audits should assess whether AI systems produce disparate impacts across demographic groups.

The institutional framework must establish that AI systems are research and analytical tools, not decision-makers. Final judgment must remain exclusively human, with judges required to articulate their reasoning in ways that demonstrate independent analysis rather than algorithmic deference. Because AI systems in justice effectively shape legal outcomes, their adoption should be subject to public oversight, legislative input, and judicial rulemaking processes that allow for democratic accountability. Procedural rules must also be adapted to the new circumstances.

12 REDEFINING INDEPENDENCE FOR THE ALGORITHMIC AGE

This institutional approach requires us to recalibrate our conception of judicial independence. Independence has never been absolute; judges operate within procedural rules, evidentiary standards, and ethical codes designed to ensure fairness and consistency. Restricting individual judges from independently adopting AI systems is not an assault on judicial independence but rather an extension of existing frameworks that delimit discretion to protect litigants and the integrity of the justice system.

Judicial independence in the AI age will require protection from technological capture. When judges adopt AI tools without institutional support or oversight, they risk becoming dependent on systems they don't fully understand, accountable to commercial vendors rather than legal principles, and unwitting participants in the erosion of the very values their independence is meant to safeguard.

The alternative, a justice system fractured by technological inequality, compromised by unexamined algorithmic bias, and rendered unpredictable by idiosyncratic AI adoption, represents not independence but chaos. It transforms the noble concept of judicial discretion into a justification for systematised arbitrariness.

13 THE EVOLVING ROLE OF THE JUDGE: FROM ADJUDICATOR TO MANAGER AND AUDITOR

The traditional image of a judge as a passive arbiter who merely listens and pronounces judgment is becoming obsolete. AI necessitates a more dynamic, proactive, and managerial role. Especially since the intrusion of AI in the judicial domain, new imperatives dictate the development of juridical competences.

In the first place, AI literacy is no longer optional; it is now part of judges' professional responsibility. A judge who is ignorant of AI is like a judge who does not understand forensic science, vulnerable to manipulation and unable to critically assess the evidence presented. Judges must understand what a specific AI tool is designed to do. Interpreting the reasoning behind AI system outputs requires a literate judge to know the right questions to ask: What data was this model trained on? How has its accuracy been validated? Can the provider explain, in understandable terms, the factors that led to this specific output? This allows them to fulfil their duty to give reasons for their decisions, even when assisted by AI.

Safeguarding procedural fairness means AI literacy enables judges to ensure that the use of technology does not violate parties' rights to a fair trial, including the right to be informed of and challenge evidence used against them, even when that evidence is an algorithmic recommendation.

14 THE INEQUALITY RISK

An AI-literate judge will be in a different position than a judge who, for educational or geographic reasons, lacks access to AI systems. The first one could analyse thousands of cases and identify pertinent precedents based on extensive data. In contrast, the second, without such tools, relies on traditional methods and personal experience. Thus, if no commonly available AI systems are in place, two identical cases could be treated differently in sheer inequality.

The inequality deepens when we consider access across socioeconomic lines. Wealthier jurisdictions may provide their judges with superior AI tools, while rural or underfunded courts lag. Private commercial AI systems create a two-tier judiciary: those who can afford premium tools and those who cannot. Even within a single courthouse, disparities emerge based on individual judges' technical literacy, personal finances, or willingness to adopt new technologies, not to mention inequalities in lawyers' ability to provide AI-based solutions, which depend on their financial capacity.

15 BIAS AMPLIFICATION AND THE EROSION OF DUE PROCESS

Risks and concerns extend beyond inequality of access to the nature of AI systems themselves. Machine learning algorithms are trained on historical data, which may, in some cases, incorporate systemic biases.²²²³that are likely to be mitigated if the reasoning were orchestrated by a judge in non-AI circumstances, but can resurface in an unprocessed data pool whose relevance is judged by AI systems²⁴. When individual judges adopt AI systems without centralised screening, they risk importing these biases directly into their judgments. An AI tool trained on biased historical data will recommend inappropriate treatment for demographic groups that have historically been discriminated against. On the other hand, a legal research AI trained predominantly on “over-purified” cases might fail to surface precedents relevant to novel circumstances.

Crucially, ad hoc use of AI systems means all kinds of biases feed into the official system. Any system without the systematic review, validation, or transparency that institutional deployment would require gradually becomes polluted. Equally, privately managed systems likely lack audit trails, are not tested for disparate impact, and have no mechanism for defendants to challenge the tool's use or to interrogate its recommendations. The judge's independence thus paradoxically obstructs the detection of algorithmic discrimination.

This undermines procedural justice, the right to understand and challenge the basis of judicial decisions. When AI systems influence judicial reasoning invisibly, defendants are denied meaningful participation in their own cases. They cannot cross-examine an algorithm. They cannot appeal a sentence if they never knew it was used, even if the AI tool was flawed. Maybe one paramount ethical standard should be that judges who employ AI systems beyond officially designated ones should disclose such use.

16 PSYCHOLOGICAL DIMENSIONS: OVERRELIANCE ON AI

Psychological research suggests that overreliance on AI by judges may stem from cognitive biases, trust dynamics, and decision fatigue, potentially undermining judicial independence and moral reasoning. Automation bias leads judges to favour AI-generated recommendations even when they conflict with their own judgment, especially under time pressure or cognitive load, leading to uncritical acceptance of AI outputs.

Cognitive offloading means that AI systems can serve as mental shortcuts, allowing judges to delegate complex reasoning. While this reduces mental strain, it may also erode deliberative depth and moral engagement. Excessive trust in technology can lead judges to perceive AI as impartial and data-driven, fostering excessive trust. This can obscure the fact that algorithms often encode historical biases and lack contextual sensitivity.

Reduced moral engagement occurs when AI handles morally charged decisions (e.g., bail or sentencing), and judges may experience moral disengagement, treating these decisions as technical rather than ethical. The presence of AI alters the symbolic meaning of judging, from a human-centred act of justice to a procedural task. This shift can affect how judges perceive their role and how the public perceives judicial legitimacy.

Under heavy caseloads, judges may increasingly rely on AI to conserve cognitive resources. This reliance can become habitual, diminishing critical scrutiny and reinforcing systemic biases. Loss of discretion means psychological reliance on AI may lead judges to defer to algorithmic outputs even when they conflict with legal nuance or ethical judgment. Research shows that while some groups may view AI-assisted decisions as fairer, others perceive them as less legitimate than human judgments.

Recommendations include training in cognitive awareness (judges should be educated about automation bias and the psychological traps of overreliance), ethical design of AI tools (systems should be transparent, explainable, and designed to support rather than replace judicial reasoning), and preserving human judgment (AI should augment, not supplant, the moral and contextual deliberation that defines justice).

17 THE RULE OF LAW AND ALGORITHMIC CONSISTENCY

The rule of law depends on predictability and consistency, the principle that similar cases should receive similar treatment under stable, publicly known standards. This doesn't mean judges cannot exercise discretion, but rather that discretion operates within a framework of shared legal principles and transparent reasoning.

Ad hoc AI adoption shatters this framework. When different judges employ different AI systems, each with its own training data, algorithmic architecture, and embedded assumptions, the factors influencing judicial decisions become radically inconsistent and uncontrollable. From the perspective of litigants and society, the law becomes incoherent, not a rational system of rules and principles, but a maze of idiosyncratic, technologically mediated outcomes that vary based on which judge happens to hear the case and which tools that judge happens to use.

Nonetheless, the rule of law requires that legal standards evolve through transparent, democratically accountable processes, legislative action, appellate review, and public debate.

This foundation is shattered by an oligopoly of AI system providers that controls legal data. When AI systems shape judicial outcomes invisibly, the interpretation of legal norms shifts without public awareness or democratic input. The AI becomes a tyrant, its training data and algorithms effectively making law by influencing which precedents are found, which factors are emphasised, and which outcomes are suggested.

18 THE MANAGERIAL JUDGE IN THE DIGITAL ERA

Differential case management is a growing part of a judge's function, and AI supercharges this aspect. In a modern context, judges are involved in docket management and triage. Even more, AI systems can enhance these functions by analysing case files to prioritise urgent matters, suggest suitable tracks for a case (e.g., standard trial, alternative dispute resolution), and even plan case duration. The judge's role evolves to overseeing this system, validating its recommendations, and making the final managerial call.

In parallel, judges obtain enhanced process control. AI-driven analytics enable judges to monitor the progress of hundreds of cases, identify bottlenecks, and ensure the efficient administration of justice across their entire docket, not just in individual cases.

19 NEW PUBLIC MANAGEMENT: AI AS A MANAGERIAL REFORM TOOL

New Public Management (NPM) reframes public institutions through the lens of private-sector efficiency, performance metrics, and customer orientation. In this context, AI becomes a transformative tool aligned with NPM principles: performance measurement (AI enables data-driven tracking of judicial efficiency, case resolution times, and workload distribution); cost reduction (automation of routine tasks reduces administrative overhead and resource consumption); standardisation and accountability (algorithms enforce consistent decision-making, which NPM views as a safeguard against arbitrariness); and user-centric services (AI-driven platforms can improve access to justice by streamlining procedures and enhancing transparency).

However, this managerial transformation is not without risks. NPM's emphasis on efficiency may conflict with the judiciary's deeper mission of justice, fairness, and deliberation. When AI is deployed primarily to meet performance targets, it risks becoming a tool of managerial domination, sidelining ethical and civic concerns.

20 CONTROL OVER AI IMPLEMENTATION: THE CORE OF INDEPENDENCE

Effective judicial independence in an AI environment requires that the tools judges rely on are neither controlled by the executive branch nor by private vendors. Therefore, the most significant element of the judicial role nowadays is the responsibility to maintain sovereign control over the technological infrastructure of justice.

AI will not replace judges, but judges who understand and shape AI will take precedence over those who do not. The preservation of judicial independence in the age of AI demands more than caution; it requires proactive engagement, institutional adaptation, and a reimagined judicial identity that encompasses technological stewardship. By embracing AI literacy, asserting control over data and system design, and expanding their roles into governance, judges can ensure that AI enhances, rather than erodes, their independence and the legitimacy of the justice system as a whole.

The future of fair and impartial justice depends not on resisting technology, but on judges leading its ethical integration. The judiciary cannot outsource its core functions. Judges, through their judicial councils or governing bodies, must be at the forefront of defining the functional requirements for AI systems. The judiciary itself must develop stringent ethical and technical standards for any AI tool used in courtrooms. This includes mandates for transparency, auditability, anti-bias testing, and protection of fundamental rights.

Vendor scrutiny and contractual safeguards mean procurement contracts must be negotiated from a position of strength, ensuring that the judiciary retains ownership of its data, has a right to audit the algorithms, and that vendors are liable for defects that lead to miscarriages of justice.

21 DATASET DEVELOPMENT AND CONTROL: OF PARAMOUNT IMPORTANCE

Perhaps the most critical dimension of judicial control lies in the development and curation of datasets used to train AI systems. Historical legal data often encodes past injustices; uncritically using such data risks automating discrimination. Judges, through judicial councils, ethics committees, or specialised AI oversight bodies, must have a decisive voice in data selection and preprocessing.

Fundamental issues arise in determining which cases, rulings, or administrative records are appropriate for training models and which must be excluded later. Their role is also essential for bias auditing via regular audits. AI models must also guarantee transparency and explainability. Judges must insist on interpretable models and accessible documentation so that parties can challenge algorithmic inputs. If left without judicial involvement, data control will be in the hands of technical corporations or dictated by commercial, legislative, or executive choices.

An AI model is fundamentally a product of its data. Biased, incomplete, or unrepresentative data will produce biased and unjust outcomes. Therefore, the development and curation of datasets are foundational judicial responsibilities. Unprocessed historical judicial data often reflect past societal and judicial biases, outdated solutions, or even an abolished procedural framework. Using this data to train a predictive model would automate and perpetuate these biases or errors. Judges must oversee the creation of curated, balanced, and ethically sourced datasets that are fit for purpose.

A judicial public good means that high-quality, anonymised legal datasets should be treated as a critical public good, developed and maintained by or under the authority of the judiciary. This prevents reliance on proprietary datasets from private companies, which can lead to lock-in and hidden bias.

Ongoing auditing and maintenance recognise that datasets are not static. Judges must ensure there are processes for continuously auditing data and AI performance for emerging biases or inaccuracies.

22 REWORKING BIASED DATASETS IN JUDICIAL AI SYSTEMS

Courts must rework datasets known to introduce bias through a structured, multi-stage process that aligns with fairness principles in AI governance. This involves identifying biases and applying debiasing techniques throughout the data lifecycle. The goal is to create representative, error-free datasets that minimise disparate impacts while preserving utility for judicial tasks such as risk assessment or case prediction.

Key methods include pre-processing techniques (modify the dataset before model training to address imbalances through resampling, reweighting samples based on sensitive attributes, or removing disparate impacts through data cleansing); in-processing techniques (integrate fairness constraints during algorithm training, such as adjusting the loss function to optimise for both accuracy and fairness metrics); and post-processing techniques (adjust model outputs after training to enforce fairness, such as recalibrating predictions for different demographic groups).

Courts should involve multidisciplinary teams (e.g., data scientists, ethicists, and judges) and use open-source toolkits like IBM's AI Fairness 360 or Microsoft's Fairlearn for implementation. EU regulations, such as the AI Act, mandate such quality evaluations for high-risk systems, requiring datasets to be relevant, representative, and error-free. Certification of reworked datasets by independent bodies can ensure compliance.

23 LIABILITY FOR FAULTY DATASETS

In cases of faulty datasets leading to biased judicial AI outcomes, liability can be recognised under several legal frameworks, focusing on harms like discrimination or unjust decisions. This typically falls under tort law, civil rights statutes, or product liability, without requiring proof of intent. A particular issue affected the EU domain following the withdrawal of the proposed AI liability directive.

Disparate impact liability under civil rights laws means entities can be liable if datasets cause disproportionate harm to protected groups, even if unintentional. Plaintiffs must show statistical disparities, after which the defendant proves a legitimate interest, and alternatives are considered in judicial contexts.

Negligence and duty of care mean developers, deployers, or courts may be liable if they fail to mitigate foreseeable biases, based on a calculus weighing the burden of prevention against the probability and severity of harm. This extends along the data lifecycle to data collectors, processors, and users.

Product liability means AI vendors can face strict liability for defective products if faulty datasets render tools unsafe, as in discrimination lawsuits against biased algorithms. Contracts may shift risks, but courts increasingly hold vendors accountable. Under the EU AI Act, providers of high-risk judicial AI face fines for non-compliance with dataset quality requirements.

Liability often involves class actions for group harms, with remedies including damages, injunctions, or system reforms. Judicial ethics rules may also impose sanctions on judges using biased AI.

The correct operation of a corrected dataset must be iteratively tested through rigorous fairness audits and metrics, ensuring it no longer amplifies biases while maintaining accuracy, especially in the case of judicial AI, where pertinent data reorganise as case law evolves. Auditing and statistical methods should be used to detect inconsistencies, biases, or imbalances compared to hypothetical human decisions. Risk-based audits under the AI Act, with human oversight, including Data Protection Impact Assessments, must also be conducted and, if necessary, independent third-party certification for high-risk systems.

24 REASSESSING CASES JUDGED UNDER A FAULTY DATASET

Justice systems should reassess cases decided on faulty datasets to uphold due process, notably when biases led to prejudicial outcomes. Novel procedural provisions must be introduced to warrant mandatory reassessment when new evidence of faults emerges. Grounds for reassessment should include due-process violations arising when AI lacks transparency or when parties cannot challenge it, inverting the burden of proof. Retrospective audits can identify affected cases.

25 CANADIAN JUDICIAL COUNCIL GUIDELINES: A MODEL APPROACH

A good example of an appropriate policy to ensure judicial independence vis-à-vis AI challenges is the Guidelines drafted by the Canadian Judicial Council (CJC), the federal agency established under the Judges Act to supervise Canada's federally appointed judges and to maintain and improve the quality of judicial services in Canada's superior courts. The CJC recently released the first edition of its Guidelines on the Use of Artificial Intelligence in Canadian Courts, published in September 2024.²⁵

The Guidelines recognise that judges are exclusively responsible for their judicial decisions and that they are not permitted to delegate their decision-making authority, whether to a judicial assistant, an administrative assistant, or an artificial intelligence computer program, regardless of the abilities of the person or program. That said, the Guidelines also recognise that some judges have already adopted AI tools to improve efficiency and accuracy, while others may rely on them without even realising it.

The objective of the Guidelines is therefore to raise awareness of the risks associated with the use of any form of AI in court administration and judicial decision-making and to prevent the delegation of decision-making authority while encouraging the safe, effective and appropriate use of AI by the judiciary.

As for protecting the independence of the judiciary, the CJC considers that "many generative AI applications proposed for use in the courts, including case management systems and alternative dispute resolution, could erode judges' authority and independence. Too much reliance on private AI (whether commercial or publicly funded) could undermine the independence of the judiciary. It must be understood that the role of judges goes beyond resolving individual disputes: they have the crucial task of interpreting the law, changing it, and constituting a stable and distinct third branch of government. Even as governments advance laws governing AI use, the independence of the judiciary must be preserved."²⁶

The Guidelines emphasise that any review of judges' use of assistive AI technology must always be consistent with the core values of independence, integrity and respect; diligence and competence; equality and impartiality; fairness; transparency; accessibility; timeliness; and certainty.

The standardisation of rulings through algorithmic systems risks reducing the judiciary to a mechanistic process, stripping away the nuance and discretion essential to justice. Such standardisation reflects broader political and managerial imperatives, efficiency, predictability, and control, at the expense of human-centred adjudication.

AI systems often present their outputs as objective and neutral, masking the value-laden choices embedded in their design.

This erosion of pluralism is particularly dangerous in legal contexts, where legitimacy depends on balancing competing values. A judiciary governed by algorithms risks becoming a technocratic apparatus, alienated from the social and ethical context it is meant to serve. This creates a crisis of justification, where judicial decisions may be procedurally sound but lack moral legitimacy.

Judges, administrators, and policymakers may support AI adoption not solely for its merits but because it aligns with institutional incentives, reducing caseloads, minimising accountability, and securing funding or political favour.

Judicial incentives mean AI can shield judges from criticism by outsourcing controversial decisions to "neutral" algorithms. In terms of bureaucratic expansion, agencies may promote AI to justify budget increases or expand their influence. In parallel, private tech firms may lobby for AI integration, shaping legal standards to suit proprietary interests.

This perspective challenges the assumption that AI adoption is purely rational or benevolent. Instead, it reveals how institutional actors may pursue AI for reasons that conflict with public interest or democratic accountability.

26 CONCLUSION: TOWARDS ALGORITHMIC SOVEREIGNTY

To safeguard judicial independence in the age of AI, justice systems must evolve towards a model of “algorithmic sovereignty.” The challenge is not merely to protect judicial independence from AI, but to redefine it, expanding its scope to include autonomy from algorithmic and technological control.

This implies ensuring that courts retain control not only over legal reasoning but also over the digital infrastructures that mediate it. Concrete measures include: embedding 'human-in-command' principles in all AI-assisted decision systems; developing publicly owned, open-source judicial data infrastructures; enhancing algorithmic literacy among judges and court staff; establishing judicial ethics committees to oversee AI deployment; and embedding transparency-by-design requirements in all justice-related AI applications.

AI's integration into judicial systems presents a paradox. Properly governed, AI can enhance judicial quality and transparency; poorly managed, it risks hollowing out the very autonomy it promises to support. On the one hand, Artificial Intelligence promises to revolutionise judicial systems by enhancing efficiency, consistency, and access to justice. Positively, AI can free judges from administrative burdens, allowing them to focus on core judicial reasoning.

The same technology that promises to enhance judicial efficiency simultaneously threatens the independence that legitimates judicial authority, since it introduces risks of "algorithmic bias," opacity in decision-making, and the potential delegation of judicial discretion to unaccountable private corporations. The central challenge is to harness AI's benefits while safeguarding judicial independence from technological erosion.

Judicial independence exists not for judges' benefit but for those who come before the courts seeking equal treatment under the law. In the algorithmic age, protecting that equality requires insisting that AI systems enter courtrooms only through institutional channels that ensure validation, standardisation, transparency, and accountability.

The alternative to an independent judiciary risks falling into randomness or arbitrariness. If we want to avoid fracturing the judicial system, AI in courts must not evolve into a tool individual judges can adopt at will, but rather into a systemic intervention requiring systematic control.

A balanced approach that combines human judgment with AI tools is necessary to enhance the efficiency of the justice system while preserving judicial independence and protecting human rights. The integration of artificial intelligence into judicial systems has sparked intense debate across legal, technological, and ethical domains.

The danger is not that AI will replace judges tomorrow, but that incremental adoption will gradually reconfigure the judicial role until independence becomes a hollow formalism. Judges may retain the aura and reputation of independent arbiters while losing the substantive capacity to reason, distinguish, and decide autonomously.

The future of judicial independence thus lies in legal systems' capacity to reconcile tradition with technological governance. The solution is not the rejection of AI but its constitutional containment. AI's emergence in judicial systems represents both a transformation and a test for the age-old principle of judicial independence. Judicial independence must be reimagined for the digital age, not as merely freedom from political interference, but as freedom from technological dependence.

As courts become data-driven institutions, independence must evolve to encompass control over digital processes, algorithms, and infrastructures. However, to prevent AI from becoming a threat to judicial independence, the role of the judge must fundamentally expand. This requires that judges themselves, through their representative bodies, assert control over AI adoption, training, and governance. Judges must become AI-literate managers and vigilant auditors of technology. Most critically, the judiciary as an institution must assert control over the implementation of AI and the development of the datasets that underpin it.

By imposing dataset governance responsibilities that judges are ill-equipped to handle, reframing precedent as statistical correlation, and transforming adjudicators into algorithmic supervisors, AI risks reducing judges from constitutional guardians to "mere editors" of machine outputs.

Appropriate judicial governance can ensure that AI will not become an unaccountable master that compromises the impartial and independent administration of the law. The challenge is not to reject AI, but to embed it within a judiciary that is pluralistic, accountable, and ethically grounded.

The preservation of judicial independence in the AI era requires proactive engagement, institutional adaptation, and a reimagined judicial identity that encompasses technological stewardship. By embracing AI literacy, asserting control over data and system design, and expanding their roles into governance, judges can ensure that AI enhances, rather than erodes, their independence and the legitimacy of the justice system as a whole.

The future of fair and impartial justice depends not on resisting technology, but on judges leading its ethical integration. AI will not replace judges, but judges who understand and shape AI will replace those who do not.

FOOTNOTES

REFERENCES

AUTHORS INFORMATION

RIGHTS AND PERMISSIONS

EDITORS

ABOUT THIS ARTICLE

DETAILS FOR PUBLICATION

AI DISCLOSURE STATEMENT

АНОТАЦІЯ