AI Transparency in Trustworthy AI : From Metaphor to Governance Tool in EU Technology Regulation
Author
Summary, in English
Transparency has emerged as a fundamental component of ethical AI guidelines around the world. In the European Union (EU), it is recognised as one of the core principles for fostering Trustworthy AI, and serves as a cornerstone in building an ecosystem of trust within the AI governance framework.
However, to support these ambitious policy objectives, the concept of transparency must be translated into clearly defined and implementable measures. Thus, by employing a combination of legal-doctrinal and socio-legal approaches, this compilation thesis aims to contribute to a clarified understanding of the concept of AI transparency in the EU’s AI governance discourses. I examine the concept of AI transparency across four levels of abstraction: as a stand-alone objective, as a governance ideal, as a governance tool, and as a ‘floating signifier’. Focusing in particular on AI transparency as a governance ideal and as a governance tool in relation to the EU’s policymaking objective of Trustworthy AI, I analyse how AI transparency has been conceptualised, designed, and implemented for two stakeholder groups — individuals and oversight bodies — within the governance frameworks of the General Data Protection Regulation (GDPR), the Digital Services Act (DSA), and the Artificial Intelligence Act (AIA).
The main argument advanced in the thesis is that while AI transparency directed at individuals (understood as data subjects, service recipients, and natural persons) remains relevant, it is the effectiveness of oversight-oriented AI transparency that is crucial to the enforcement of the EU technology regulation and is, ultimately, foundational in the EU’s pursuit of Trustworthy AI. Although transparency is central to the EU’s vision for Trustworthy AI, its effectiveness depends on how legal obligations are interpreted, implemented, and enforced in practice.
However, to support these ambitious policy objectives, the concept of transparency must be translated into clearly defined and implementable measures. Thus, by employing a combination of legal-doctrinal and socio-legal approaches, this compilation thesis aims to contribute to a clarified understanding of the concept of AI transparency in the EU’s AI governance discourses. I examine the concept of AI transparency across four levels of abstraction: as a stand-alone objective, as a governance ideal, as a governance tool, and as a ‘floating signifier’. Focusing in particular on AI transparency as a governance ideal and as a governance tool in relation to the EU’s policymaking objective of Trustworthy AI, I analyse how AI transparency has been conceptualised, designed, and implemented for two stakeholder groups — individuals and oversight bodies — within the governance frameworks of the General Data Protection Regulation (GDPR), the Digital Services Act (DSA), and the Artificial Intelligence Act (AIA).
The main argument advanced in the thesis is that while AI transparency directed at individuals (understood as data subjects, service recipients, and natural persons) remains relevant, it is the effectiveness of oversight-oriented AI transparency that is crucial to the enforcement of the EU technology regulation and is, ultimately, foundational in the EU’s pursuit of Trustworthy AI. Although transparency is central to the EU’s vision for Trustworthy AI, its effectiveness depends on how legal obligations are interpreted, implemented, and enforced in practice.
Publishing year
2025-06-10
Language
English
Full text
Document type
Dissertation
Publisher
Department of Technology and Society, Lund University
Topic
- Law
- Artificial Intelligence
Keywords
- Transparency
- Artificial intelligence
- AI
- AI transparency
- Trustworthy AI
- European Union
- EU
- AI Act
- Digital Services Act
- General Data Protection Regulation
- GDPR
Status
Published
Research group
- AI and Society
Supervisor
ISBN/ISSN/Other
- ISBN: 978-91-8104-549-9
- ISBN: 978-91-8104-550-5
Defence date
10 June 2025
Defence time
09:15
Defence place
Lecture Hall E:1406, building E, Klas Anshelms väg 10, Faculty of Engineering LTH, Lund University, Lund.
Opponent
- Lena Enqvist (Assoc. Prof.)