🤖⚖️ CJEU to Rule on AI‑Assisted Expert Evidence Under the AI Act...
A new CJEU reference (C-245/25) examines whether AI-assisted accident-simulation software used by court-appointed experts qualifies as a high-risk AI system under the AI Act. The case raises fundamental questions about human oversight, explainability, and the admissibility of AI-generated expert evidence in judicial decision-making.
🚗 The case originates from a road-traffic accident and the question of who was at fault. A forensic accident investigator was appointed to assess the situation and used the software “Virtual Crash 4.0” to support the analysis. This version was designed for the US market (with US standards, road conditions, etc.), leading the applicant to argue that the tool was inappropriate. The applicant also insisted that the expert report should include a full description of all calculations, but the expert could only confirm that he had verified the program’s output.
📘 The Bulgarian court considered that a preliminary ruling from the CJEU was needed. It referred, among other things, to the Charter, Opinion No. 26 (2023) of the CCJE, the CJEU’s AI Strategy, the principle of explainability, and the obligation for courts to understand the decision-making process.
❓ The court submitted several questions going to the core of the AI Act and automated decision-making. Is the software a high-risk AI system within the meaning of point 8 of Annex III of the AI Act? What is the scope of Article 86 AI Act on the right to an explanation of individual decision-making? What role does Article 19 TEU (fair trial) play, and can a court rely on an expert report based on AI-generated output that has been “verified” only through the expert’s professional experience and knowledge?
💭 Not sure whether the CJEU will answer all these questions. After all, it all starts with the threshold question: is Virtual Crash 4.0 even an AI system?