Integration of AI into the Justice System of Ukraine: Normative Boundaries, Technological Sovereignty, and Case-Law

AGENDA

• 1. European and international context
• 2. National normative and ethical framework
• 3. Institutional policies, cybersecurity, and risk management
• 4. Practical tools and digital infrastructure
• 5. Case-law: the limits of permissible use of AI

AI use among the population of Ukraine

• Over the course of one year (2025–2026), the share of regular AI users in Ukraine increased from 17% to 27%. One in every six Ukrainians (17%) uses AI on a daily basis.
• Digital readiness of young people: 58% of Ukrainians aged 18–29 are regular AI users.
• This is shaping a new generation of court users who expect faster and more digitally enabled justice.

Джерела: [1]

Signing of the Council of Europe Framework Convention on AI

• Ukraine is demonstrating its commitment to European standards of justice: on 15 May 2025, Ukraine signed the Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy and the Rule of Law (CETS No. 225).
• By taking this step, Ukraine undertook to ensure that the use of AI in the public sector would not undermine the principles of judicial independence, access to justice, and protection of privacy. This lays an international legal foundation for the implementation of technologies in line with the Human-in-the-Loop principle.

Джерела: [2]

European integration benchmarks under Chapter 23 within Cluster 1: Fundamentals

• In the course of Ukraine’s EU accession process, the justice system is actively implementing the codified benchmarks of Chapter 23 “Judiciary and Fundamental Rights”, communicated in February 2026.
• The EU’s codified benchmarks for Chapter 23 specifically identify the promotion of digitalisation of the justice system (IBM 23.4.5) and the improvement of case allocation (IBM 23.2.6), thereby establishing for Ukraine a framework of institutional efficiency, accountability, and proper administration of justice.

Джерела: [3]

State strategy: Action Plan for the regulation of AI

• A key policy instrument for AI development in Ukraine is Cabinet of Ministers of Ukraine Order No. 457-r dated 9 May 2025, which approved the Action Plan for the implementation of the Concept for the Development of Artificial Intelligence.
• This document assigns a specific task to the Ministry of Digital Transformation to develop and submit a comprehensive legislative framework for the legal regulation of AI in Ukraine by the end of 2026, aligning the national approach with the European AI Act.

Джерела: [4]

Code of Judicial Ethics (Article 16): permitted and prohibited uses of AI

• Ukraine became one of the first countries to enshrine the use of AI in judicial ethical standards.
• In September 2024, the Congress of Judges of Ukraine introduced a new Article 16 into the Code of Judicial Ethics.
• This Article defines AI as an exclusively auxiliary tool. Its use is permissible only insofar as it does not affect judicial independence or impartiality, does not involve the evaluation of evidence, does not interfere with final decision-making, and does not violate legal requirements.

Джерела: [5]

Commentary on the Code of Judicial Ethics regarding AI (2026).

• To clarify the new ethical boundaries, on 2 March 2026 the Council of Judges of Ukraine (Decision No. 14) officially approved a detailed Commentary on the Code of Judicial Ethics.
• The Commentary establishes strict prohibitions: AI may not suggest the outcome of a case, carry out the legal assessment of facts, or prepare the reasoning part of a judgment without effective judicial control. Only the systematisation of materials, case-law research, and linguistic analysis of text are permitted.

Джерела: [6]

Guidance for legal professionals (Ministry of Digital Transformation and Ministry of Justice)

• In July 2025, the Ministry of Digital Transformation jointly with the Ministry of Justice developed specialised “Recommendations on the Responsible Use of Artificial Intelligence for Legal Professionals”.
• The document incorporates the principle of protection of legal professional privilege and commercial secrecy, expressly advising against entering clients’ personal data into open generative models due to the risks of secondary use.
• This is an important step towards building an ecosystem of responsible AI use in justice.

Джерела: [7]

Regulation on the Use of AI Technologies by Employees of the Office of the Supreme Court

• On 8 December 2025, the Regulation on the Use of AI Technologies by Employees of the Office of the Supreme Court (Order No. 117) was approved, comprehensively regulating the processes of integrating innovation within the Supreme Court.
• The Regulation defines AI as an auxiliary tool only, requires professional verification of all AI-generated outputs, and places full personal responsibility on the employee. It prohibits the use of public AI systems for restricted-access information, automatic drafting of judicial acts, attempts to predict individual judicial decisions, and the monitoring of employees’ behaviour.

Джерела: [8]

Institutional rules of the High Anti-Corruption Court (HACC)

• In December 2024, the HACC approved its “Principles for the Use of AI Tools”, becoming the first court to adapt data management technologies in view of the high risks inherent in its jurisdiction.
• The document expressly prohibits the uploading of any official court documents to publicly accessible cloud-based AI systems, while permitting the use of algorithms exclusively for the optimisation of routine work, the visualisation of statistical data, and communication management.

Джерела: [9]

HACC Development Strategy for 2026–2028

• In November 2025, the HACC approved its Development Strategy, which identified AI integration as a key driver of growth in the court’s operational efficiency (Decision of the Judges’ Meeting No. 1).
• The Strategy places particular emphasis on the creation of automated hearing transcription tools, the implementation of secure IT solutions, and the development of an internal crisis response protocol for cyber threats and possible leaks of confidential data through digital channels.

Джерела: [10]

National cybersecurity standards for AI systems (2026)

• In February 2026, the State Service of Special Communications and Information Protection (Order No. 154) approved the first Recommendations on the Cyber Protection of Systems Using AI.
• These Recommendations provide protection for courts against specific threats: data poisoning and prompt injection. They provide for the implementation of differential privacy safeguards, ensuring the protection of judicial data sets during machine learning.

Джерела: [11]

Concept for the development of the USICS: AI in the architecture of judicial proceedings

• Approved in April 2025, the new Concept of the Unified Judicial Information and Communication System (USICS) provides for the deep integration of AI into the life cycle of the electronic case file.
• AI modules will be responsible for the automatic recognition and classification of documents, the depersonalisation of judicial decisions prior to their publication (data masking), semantic search for relevant case-law, and even the creation of AI-powered virtual assistants for enhancing access to justice.

Джерела: [12]

Modernisation of the Supreme Court Case-law Database 2.0

• A vivid example of practical implementation is the updated version of the Supreme Court Case-law Database (LPD 2.0), presented in spring 2025.
• This national database has integrated generative AI components. The system is now capable of instantly processing vast arrays of case-law, generating concise, structured summaries of legal positions for users and identifying inconsistencies in judicial practice.

Джерела: [13] [14]

Pilot project of the Cabinet of Ministers of Ukraine: AI assistant for administrative offences

• The Government of Ukraine plans launch an experimental project using AI for the handling of high-volume administrative offence cases (in particular, traffic violations).
• The system, operating under the Human-in-the-Loop principle, is designed to analyse police reports using automated image and text recognition tools, detect errors, and prepare a draft ruling. At the same time, AI never makes a decision independently — final control and verification always remain within the exclusive competence of the judge.

Джерела: [15]

Development of “Sovereign AI” and the national Diia AI model

• Ukraine has begun building the infrastructure of “Sovereign AI” in partnership with NVIDIA in order to create its domestic computing capacity.
• In addition, a national large language model, Diia AI, is being developed on the basis of one of the open models. This will enable courts and the public sector to process official and sensitive data in a closed, secure local environment.

Джерела: [16]

Fundamental position: AI is not a source of law or an expert (Ruling of the Supreme Court dated 08.02.2024 in case No. 925/200/22)

• The Supreme Court, for the first time, categorically distinguished between the authority of a judicial decision and machine-generated output. An attempt by a participant in the proceedings to oppose ChatGPT’s conclusions to the legal position of the Supreme Court was rejected.
• The Court held that AI is an exclusively auxiliary tool and has no normative or scientific authority whatsoever. Comparing the “opinion” of an algorithm with the position of the Court was found to be impermissible and was classified as an abuse of procedural rights and as disrespect towards the justice system.

Джерела: [17]

Combating AI “hallucinations” and lawyers’ liability (Rulings of the Supreme Court dated 15.01.2026 in case No. 240/14153/24)

• In 2026, the Supreme Court considered cases in which participants in proceedings filed cassation appeals containing references to non-existent (“AI-fabricated”) precedents and case numbers.
• The Court held that the submission of AI “hallucinations” without mandatory meaningful human oversight (human-in-the-loop) constitutes a gross violation of the principle of good faith. Responsibility for any AI-generated content submitted to a court lies exclusively with the person who submits it.

Джерела: [18]

Prohibition on substituting judicial reasoning with AI-generated text (Ruling of the Kyiv Court of Appeal dated 30.07.2025 in case No. 11-кп/824/1818/2025)

• The case reflects a strict judicial approach to the drafting of judicial decisions or parts thereof with the aid of AI. The Court of Appeal quashed the judgment of the court of first instance because its reasoning section had been generated by AI.
• The Court emphasised that the intellectual evaluation of facts and the interpretation of legal norms are an inherent competence of a human judge. Delegating this function to an algorithm is unlawful and nullifies the basic standards of a fair trial.

Джерела: [19]

Rejection of AI analysis as procedural evidence (Judgment of the Supreme Court dated 08.07.2025 in case No. 925/496/24)

• A party to the proceedings attempted to use the results of AI-based analysis of the text of a contract as proper evidence supporting its position in a property dispute.
• The Supreme Court rejected such attempts, indicating that AI-based software is not recognised as an independent source of evidence and cannot provide conclusions having evidentiary force. AI-based argumentation is treated by the court solely as the party’s own submissions, which do not substitute for judicial expert examination.

Джерела: [20]

Protection of privacy: blocking attempts at AI de-anonymisation in criminal proceedings (Ruling of the Appeals Chamber of the HACC dated 01.09.2025 in case No. 991/3222/25)

• The appellant attempted to prove her own guilt on the ground that AI had managed to “de-anonymise” her identity by position, by correlating the text of a judgment (concerning another person under a plea agreement) with open-source data.
• The Court acted to protect privacy and rejected the AI “conclusions”, stating that algorithmic correlations (mosaic re-identification) have no evidentiary or prejudicial force, and that AI assumptions cannot automatically establish legal facts concerning individuals.

Джерела: [21]

Procedural time limits are not suspended due to the use of AI (Ruling of the Supreme Court dated 19.06.2025, case No. 520/6119/23)

• The appellant missed the time limit for appeal and sought its renewal, arguing that it had been necessary to use a paid version of AI in order to formulate a legal position in a complex case.
• The Supreme Court held that the use of paid AI versions or the time spent waiting for AI analysis is a subjective problem of the lawyer. This is not recognised as an objective insurmountable circumstance, since professional legal defence must be based on the lawyer’s knowledge, rather than on technology.

Джерела: [22]

Assessment of a judge’s impartiality is not subject to algorithmic analysis (Ruling of the Supreme Court dated 12.09.2025 in case No. 750/6682/23)

• The applicant sought the recusal of a panel of judges, relying in support of the motion on “the purportedly ‘impartial opinion’ of AI”, which had analysed the judges’ conduct and allegedly confirmed their bias.
• The Supreme Court categorically rejected such an approach. The Court stated that objectivity and impartiality are ethical and legal categories, and that a machine algorithm is incapable of assessing the human or moral dimension of a judge’s work. AI conclusions have no evidentiary significance in this context.

Джерела: [23]

PREVIOUS RESEARCH AND AUTHOR’S CONTRIBUTIONS

• Bernaziuk Ian. Artificial intelligence in the Ukrainian judiciary: charting the course under the digital gavel https://court.gov.ua/eng/supreme/pres-centr/news/1891488
• Bernaziuk Ian. Artificial Intelligence and The Judicial System Of Ukraine: Results Of Cooperation In The Past Year https://court.gov.ua/storage/portal/supreme/prezentacii_2025/AI_Ukraine_bernaziuk.pdf
• Bernaziuk Ian. Benchmarking Justice: Can AI Uphold the Rule of Law? (20.11.25) https://court.gov.ua/storage/portal/supreme/prezentacii_2025/156_AI_Benchmarking_Justice_bernaziuk.pdf
• Bernaziuk Ian. Artificial Intelligence in the Ukrainian Judiciary: Charting the Course Under the Digital Gavel https://constitutionalist.com.ua/artificial-intelligence-in-the-ukrainian-judiciary-charting-the-course-under-the-digital-gavel
• Bernaziuk Yan, Barikova Anna. Discretion in Applying Provisions of Law: Linguistic Prospects for AI and Machine Learning. ECS Transactions, 2022, 107(1), pp. 18545–18558 https://iopscience.iop.org/article/10.1149/10701.18545ecst

Джерела: [24] [25] [26] [27] [28]