The Judiciary In The Age Of Artificial Intelligence

A question once confined to works of science fiction is now knocking on the courtroom door: how, and within what limits, may judges, prosecutors, or expert witnesses make use of artificial intelligence? The fact that a commercial court in Istanbul openly acknowledged, in the reasoning of its judgment,¹ that it had made use of artificial intelligence has taken this question out of the theoretical realm and turned it into a live issue for the Turkish judiciary.

In the judgment, artificial intelligence was used to research foreign legal sources, verify foreign court decisions, and translate foreign-language legal texts into Turkish; the court specifically emphasized, by contrast, that the legal assessment, the weighing of evidence, the formation of judicial conviction, and the rendering of the judgment itself remained entirely the judge’s own.

The use of artificial intelligence in judicial processes can be assessed under three main categories: (i) information gathering, (ii) analysis and content generation, and (iii) verification and control. The first category may include legal research, document review, and translation; the second, the preparation of draft reports or draft decisions; and the third, the verification of sources and generated content together with consistency checks.

The question that needs to be debated today is not whether artificial intelligence should be used at all. Current developments show that excluding AI systems from judicial processes altogether is no longer a realistic approach. At the same time, using AI systems haphazardly — and, more importantly, without any oversight — may give rise to numerous problems and ultimately undermine public trust in the judiciary. What must be discussed is how, and within what limits, such use should take place, how it should be supervised, and how data security should be ensured.

Turkish law does not yet contain any regulation governing the use of artificial intelligence systems by courts and judicial personnel. For the time being, the matter can therefore only be assessed within the framework of the Public Officials’ Ethics Board’s Decision of Principle and, through interpretation, the Constitution and the rules of procedural law.

• The Public Officials’ Ethics Board’s Decision of Principle: In the reasoning of its judgment, in connection with its use of artificial intelligence systems, the court referred to the Decision of Principle issued by the Public Officials’ Ethics Board, established under Law No. 5176²,³ which sets out the fundamental ethical principles and the scope of action that public officials must observe when using artificial intelligence systems.

Under the Decision of Principle, the use of artificial intelligence directly requires human oversight. Within this framework, outputs may not be used without verification; personal data and confidential information must be protected with particular care; and due diligence must be exercised against any biased or discriminatory results the systems may produce. It is further stated that, whatever the system’s output, final responsibility remains with the user.

It should be noted, however, that under the second paragraph of Article 1 of Law No. 5176, members of the judiciary are excluded from that Law’s scope of application. Accordingly, notwithstanding the court’s reference in its reasoned judgment, decisions of the Public Officials’ Ethics Board do not carry the force of a directly binding legal norm as regards judges; they can only be regarded as an ethical and advisory recommendation.

• The Non-Delegability of Decision-Making Authority: Under Article 138 of the Constitution, judges are independent in the exercise of their duties and render judgment in accordance with the Constitution, the law, and their own conviction. This constitutional principle shows that decision-making authority belongs to the judge and cannot be delegated. It follows that an artificial intelligence system rendering the decision — or a judge adopting the system’s output without any independent assessment of his or her own — would constitute a violation of the Constitution.
• Reasoned Judgment and Transparency: Taken together, Article 141 of the Constitution and Article 27 of the Code of Civil Procedure require that the legal and factual grounds on which a court decision is based be capable of being understood. Where information or results are obtained through artificial intelligence systems, both that input and the reasoning built upon it must remain subject to review.

International Regulations on the Use of Artificial Intelligence in the Judiciary

The European Commission for the Efficiency of Justice (CEPEJ)⁴, in its European Ethical Charter on the Use of Artificial Intelligence in Judicial Systems and Their Environment⁵, has adopted the principles of respect for fundamental rights, non-discrimination, quality and security, and transparency together with human oversight. The underlying approach is that artificial intelligence must not replace the judge, and that control over the decision-making process must remain with the human being.

The European Union Artificial Intelligence Act,⁶ for its part, imposes obligations such as record-keeping, human oversight, technical documentation, and transparency on artificial intelligence systems that assist courts in researching and applying the law. Indeed, it directly classifies artificial intelligence systems used in the machinery of justice and in the conduct of judicial proceedings as ‘High-Risk AI Systems’. The Act imposes strict obligations on the use of such systems, including detailed record-keeping, a high level of human oversight, technical documentation, cybersecurity, and transparency.

For this reason, it is essential that courts establish a standardized “Artificial Intelligence Use Record” document. This record should securely and traceably document the artificial intelligence model and version used, the purpose of its use, the types of data transferred to the system, the prompts affecting the process, the results obtained, and the independent sources — outside the AI system itself — through which those results were verified. In this way, both the parties and the appellate courts can, where necessary, examine and review the artificial intelligence process under the principle of a fair trial.

An Assessment from the Perspective of Artificial Intelligence Systems’ Terms of Use

When using artificial intelligence systems, it is not only the AI model itself but also the provider’s terms of use that must be taken into account. For example, OpenAI, Anthropic, and Google each set out, in their published terms and conditions of use, three common core points.

• No Warranty and the Duty to Verify: No provider warrants that its outputs are legally or factually correct. On the contrary, every provider expressly states that users must independently verify the results.
• Data Security and Confidentiality: There are significant differences in data security between individual and enterprise accounts. Enterprise versions offer stronger guarantees that data will not be used to train the AI model, whereas this protection is inadequate under free individual
• Choice of System: Full responsibility for information processed by artificial intelligence systems rests with the user. For this reason, transferring case-file contents, party information, trade secrets, or the many kinds of personal data — often sensitive, often not — to a publicly available AI system poses a major risk in numerous respects, first and foremost data security.

For this reason, judges, prosecutors, and expert witnesses must use enterprise-grade systems that can be logged and audited, rather than individual subscriptions.

An Assessment of the Use of Artificial Intelligence by Expert Witnesses

The use of artificial intelligence carries significant consequences not only for judges and prosecutors but also — even though they are not members of the judiciary — for expert witnesses. Under the Expert Witness Regulation,⁷ the method used, the technical basis, and the conclusion reached must all be explained in the report. A bare statement such as “Artificial intelligence was used” should therefore not suffice; during the preparation of the report, the purpose for which the system was used, the data it worked on, how the result was arrived at, and how that result was verified must also be recorded and explained.

A notable recent development on this front has taken place in the United States,⁸ where an expert witness used artificial intelligence to filter thousands of documents; the opposing party then demanded disclosure not only of the expert’s report but also of the prompts given to the AI system.

The court held that, because the process of determining which documents would be examined directly affected the reliability of the result reached, the prompts used formed an inseparable part of the expert’s methodology.

Here, the AI system did not merely produce a final result; it also determined which pieces of evidence would be placed before the expert for examination. On that basis, the court treated the prompts given and the filtering process as auditable elements of the methodology that must, where necessary, be opened to the parties’ review.

The significance of this decision is that it may no longer be sufficient to regard artificial intelligence in litigation merely as a "supporting tool." Once AI becomes part of the method used, judicial review extends beyond the final report to how, and with what instructions, the AI was used.

Given the breadth of expert witnesses’ duties under Turkish law, and the critical role their reports play as a basis for judgments, this approach is of considerable importance. Turkey needs a dedicated piece of legislation requiring expert witnesses to record the AI prompts they use, their data inputs and outputs, and their verification processes in a manner that ensures the reliability and auditability of the report, and clarifying the principles governing such use.

By way of illustration, when we put the same question to different artificial intelligence systems — asking what kind of AI system could be used in the judiciary and how —⁹ the models’ answers diverged. Claude put forward a layered oversight model centered on constitutional safeguards, Gemini highlighted more institutional mechanisms, such as an independent “Judicial AI Ethics Board,” mandatory certification, and an AI use registry. ChatGPT placed at the center of oversight the notion of a “non-delegable judicial core,” stating that essential judicial functions such as the assessment of evidence, legal characterization, and the giving of reasons can under no circumstances be left to artificial intelligence. As this shows, quite apart from how greatly results can vary depending on the prompts or questions asked, even the very same question can produce different answers from different systems.

In conclusion, the use of artificial intelligence in judicial processes must follow a model that is human-directed, verified against independent sources, properly recorded, and subsequently open to equal and transparent review by all concerned, in keeping with the principle of a fair trial. Final legal assessment, and the responsibility that goes with it, must in every case remain with the judge, prosecutor, or expert witness.

It should come as no surprise that rapid developments are taking place, even as this note is being written, both in artificial intelligence technologies and in the legal regulations governing them. Indeed, while we were preparing this note, the Union of Turkish Bar Associations published its “Recommended Guide on the Use of Artificial Intelligence for Lawyers,” proposing a framework for the safe, informed use of AI tools in legal practice, consistent with lawyers’ professional obligations. In this fast-moving landscape, the lasting solution lies not in rigid prohibitions or stop-gap measures, but in building a legal infrastructure that can adapt to technological change — one that is sustainably flexible, agile, and auditable.

Footnotes

^{1 Istanbul 14th Civil Court of First Instance for Commercial Matters, decision dated 30.04.2026, Case No. 2025/976, Decision No. 2026/413}

^{2 Law No. 5176 on the Establishment of the Public Officials’ Ethics Board and Amending Certain Laws}

^{3 Decision of Principle No. 2024/108 dated 10.09.2024, "Ethical Principles of Conduct to Be Observed by Public Officials in the Use of Artificial Intelligence Systems"}

^{4 European Commission for the Efficiency of Justice.}

^{5 CEPEJ, European ethical Charter on the use of Artificial Intelligence in judicial systems and their environment}

^{6 European Parliament and Council. Regulation (EU) 2024/1689 (Artificial Intelligence Act). 2024.}

^{7 Expert Witness Regulation (Official Gazette No. 30143, dated 03.08.2017), Art. 55}

^{8 Conservation Law Foundation, Inc. v. Shell Oil Co., No. 3:21-cv-00933, ECF No. 970 (D. Conn. May 18, 2026)}

^{9 "What legal oversight model would you propose to ensure that the use of generative artificial intelligence in the judiciary can be supervised in a manner consistent with the rule of law, judicial independence, the right to a fair trial, and the right to a reasoned decision?"}

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.