Domstolens artificiella assistent - En analys av AI-systemens roll i Sveriges domstolar i ljuset av AI-förordningen och rätten till en rättvis rättegång

Sammanfattning

With the increasing use of artificial intelligence across various sectors of society, the Swedish judiciary has begun implementing automated solutions and tools in its operations: a redaction service intended for anonymizing confidential information and a translation service used to translate judicial decisions. However, the national legal system does not contain legislation specifically tailored to AI, which has meant that actors have had to rely on technology-neutral regulation and the regulation of fundamental rights. On 1 August 2024, an EU regulation in the field of AI technology entered into force, except for certain provisions who will become applicable in 2026. As a result to this, Sweden now has an ”AI regulatory framework” to follow. With a new legal act to follow, several questions arise. This essay addresses two main questions. The first concerns how the AI systems currently used by Swedish courts should be legally classified under the EU AI act. The second concerns how these AI systems relate to the requirements for judicial independence, with particular focus on the requirements of independent and impartial tribunals, as well as transparency, within the broader framework of the right to a fair trial. The analysis initially examines the risk classification applicable to the AI systems used by Swedish courts, based on the regulation’s system of different risk levels. This assessment is preceded by an evaluation of whether the courts, as actors, and their AI systems fall within the scope of the regulation. The analysis demonstrates that both the courts and the AI systems are covered by the regulation. It further shows that the systems do not constitute either unacceptable-risk or high-risk AI systems, as their intended purposes are purely administrative in nature and do not directly affect substantive decision making. The second part of the analysis examines the requirements of independent and impartial tribunals, as well as transparency, in relation to the AI systems used in Swedish courts, AI systems which have been found to perform only narrow procedural tasks. Based on the scope and manner of their use, the assessment concludes that the systems satisfy all three requirements. At the same time, it is emphasized that AI systems with a greater impact on legal effects than those currently in use may trigger stricter legal requirements.