The Meaning of Transparency in EU’s Regulation of AI

This article examines the concept of transparency in the regulation of artificial intelligence in the European Union, with a particular focus on the transparency requirements in the General Data Protection Regulation (GDPR) and the Artificial Intelligence Act (AIA). It further compares the legal concept of transparency in the GDPR and the AIA with transparency applied in the computer science literature. Transparency in law appears as an intuitive but also elusive concept, with no firm legal definition. Personal data shall be processed in a ‘transparent manner in relation to the data subject’ as outlined in GDPR Article 5(1)(a), and, according to AIA Article 13(1), high-risk AI systems shall be designed with ‘an appropriate type and degree’ of transparency. To better understand these and other legal transparency requirements applicable to the regulation of AI, the article offers an analytical framework of three distinct transparency concepts. Appropriate transparency, pursuant to the GDPR and the AIA, may be achieved by a combination of three types of transparency: simplicity, observation and explanation. These three distinct aspects of transparency have different qualities and should not be confused with each other.