10.29327/1884517.29-18
The European Union Artificial Intelligence Act introduces strict obligations for high-risk AI systems, yet a significant translation gap remains between its legal mandates and the precise specifications that software engineering practice demands. Article 13 in particular, which governs transparency requirements, is written at a level of abstraction that resists direct technical implementation. This paper proposes a five-phase operationalization process that bridges this gap by integrating the Non-Functional Requirements Framework with the ISO/IEC 25059:2023 quality model. Starting from the legal text, the process extracts atomic obligations, maps them to standardized quality characteristics, and pro-duces measurable acceptance criteria organized in a bidirectional traceability matrix. Applied to a credit scoring AI system, the process yields nine traceable requirements across nine quality dimensions. The analysis shows that while Article 13 can be largely operationalized through this approach, some provisions remain technically indeterminate and require organizational governance mechanisms beyond what engineering specification alone can provide. The paper contributes a structured vocabulary and a concrete roadmap for translating AI regulation into verifiable software requirements.
Keywords: EU AI Act; NFR Framework; ISO/IEC 25059; Requirements Engineering; AI Transparency; Software Quality
@inproceedings{wer202617,
author = {Almentero, E. and Sousa, H. D. S. and Portugal, R. and Classe, T. and Kalinowski, M.},
title = {Translating EU AI Act Transparency Requirements into Software Engineering Specifications},
booktitle = {Anais do Workshop em Engenharia de Requisitos - Proceedings of the 29th Workshop on Requirements Engineering (WER2026)},
year = {2026},
issn = {2675-0066},
isbn = {978-65-02-19591-8},
doi = {10.29327/1884517.29-18}
}