Levent Kennes/Stockholm
The controversial AI (AI) initiative proposed by the Turkish Ministry of Justice, known as the CBS Organizational Forecast Project, has sparked vigilance among legal experts and human rights advocates. Designed to automatically identify new case entries and potential associations between previously classified terrorist organizations and national judicial databases, the system has been criticized for potentially violating innocence and estimates of international human rights standards.
The AI tool is integrated into the Turkish National Judicial Informatics System (UYAP), which has the clear purpose of enhancing the accuracy of judicial statistics and reducing artificial errors in data entry. According to an official presentation by the Department of Justice held on April 8, during a meeting of Congressional Committees, the Department of Justice assessed the risks of public sector AI use and explored the need for legal frameworks, cross-references to the AI model entered new entry with existing databases of perceived terrorist organizations, automatically flagged discrepancies and proposed classification.
Such automation poses serious risks, particularly with regard to innocence estimation. Automatically labeling individuals or legal documents as affiliated with terrorist organizations based on algorithmic inferences could be pre-judgment. This approach risks introducing systematic bias into legal procedures and can be deemed incompatible with basic legal principles, including the right to fair trial.
CBS Organizational Forecasting Tools raise deep legal and ethical questions. By linking individuals or cases algorithmically to terrorist organisations, the system can effectively “tag” before a judicial review is undertaken. Such tagging can cause irreversible reputational harm from the start and bias the judge.
The ministry defends the project by highlighting administrative interests, including enhanced statistical accuracy and better reporting compliance with international organizations such as the Financial Action Task Force (FATF).
The CBS project is not an isolated example. The Ministry of Justice's broader AI integration strategy includes at least eight other projects that could violate international legal norms and human rights.
The Department of Justice acknowledges that AI systems are as neutral as the data used to train them. Nevertheless, the large-scale deployment of AI in decision support systems without complete transparency in data sources raises concerns. Dataset bias can have discriminatory consequences, particularly for minority groups. Without a robust mechanism to identify and mitigate bias, AI decisions risk strengthening existing bias.
The ministry argues that AI is used solely for decision support and not for final legal decisions, but that the impact of such tools on human decision makers is not trivial. If a judge relies on AI output without understanding the logic or data behind it, it compromises the legal principles of reasonable judgment. The lack of a clear explanatory mechanism contradicts the international invocation of transparent AI.
AI tools work with large databases of judicial documents. Many of them contain highly sensitive personal data. The ministry claims that data is anonymized and protected, but the risk of re-identification on large datasets remains. Data breaches or misuses can violate international standards such as the Turkey's National Data Protection Law (KVKK) and the General Data Protection Regulation (GDPR).
The deployment of AI in judicial settings raises open questions of legal liability. If the AI system contributes to a false judicial decision, it remains unclear whether the responsibility lies with the developer, state, or individual judge. This legal ambiguity could lead to the victim leaving without an effective legal remedy in violation of the right to effective remedies under Article 13 of the European Convention on Human Rights (ECHR).
Projects like the “EzCümle” initiative produce automated summaries of legal texts, and other projects that propose standard inferences in the verdict could lead to over-mechanization of the judiciary. Efficiency is improved, but there is a risk that the depth and nuance of human judicial reasoning will decrease. Such systems can unintentionally reduce justice to a procedural form, and violate the principles of human dignity and individual justice.
Some AI projects aim to predict crime trends and to predict organizational affiliation based on historical patterns. This reflects the predictive policing system that violates privacy and has been criticized in other countries for enabling racial profiling. Without strict legal protection, such practices could infringe on rights protected under Article 8 of the ECHR.
If AI has a significant impact on initial judgments or classification without clear documentation of its logic, it could undermine the validity of the appeal. Defendants will face a decision that challenges difficulties if the rationale is incorporated into an opaque algorithm and is likely to violate a fair trial guarantee.
The ministry emphasizes the existence of an internal ethics committee that oversees the implementation of AI, but the agency operates within the same management structure that develops the tools. Independent surveillance – a key component of democratic accountability – is missing, limiting transparency and increasing the risk of abuse.
At a parliamentary committee meeting, Servet Gül, Director of Information Technology, Ministry of Justice, recognized the ministry's limited technical capabilities in AI development. He said the province's artificial intelligence and big data division currently consists of only 11 technical staff. Additional administrators are involved, but he admitted that this number is insufficient. “Is that enough? No,” Gül said. Current resources remain constrained while efforts to expand the team by recruiting qualified individuals are underway. He also emphasized that AI can produce beneficial outcomes in the hands of well-intentioned people, but warned of its potential misuse if manipulated by people with malicious intent.
Dr. Osman Gazi Gütürk, a faculty member at Galatasaray University, discussed artificial intelligence. He began by displaying the official definition of AI, which was published in the official official gazette in July 2024. One of the most important changes during this period was the definition itself. “We were able to ask 20 different engineers for a review. They all tell us that the definition does not include any of the systems,” Güçlütürk said. He cautioned that by creating regulations for AI, it focused on a single type and could be overly rigid.
A significant risk in Türkiye is the security of public authorities' data. Information from institutions that frequently experience database violations are often sold on dark web marketplaces. On September 12, 2024, Transport and Infrastructure Minister Abdulkadir Uraloğlu confirmed that personal data of 85 million people had been stolen during the pandemic. The Minister acknowledged the leak from the healthcare system. However, after a fierce backlash, he insisted that there was a misunderstanding and that no data was leaked. In 2015, the Highest Election Council was hacked and voter information was stolen. In 2021, the agricultural and forestry system was also breached, losing all digital data. Opposition parties claim that the system recovered after the ransom was paid to the hackers.
Turkey ranked 117th out of 142 in the World Judicial Project's 2024 legal index, dropping one rank from the previous year. The index measures the rule of law based on factors such as corruption, fundamental rights, civil and criminal justice.
Minutes of the Parliamentary Committee to be held in Ankara on April 8th:
