European AI Act: Transparency Before, Inside, and After the AI Black Box

Following on from the successful workshops on Regulating High Risk AI in the EU (2023), and Compliance and Enforcement (2024), AI Lund's fika-to-fika workshops* returns to the AI Act with a focus on legal aspects related to “transparency”.

The EU AI Act (Regulation (EU) 2024/1689) marks a groundbreaking shift in AI regulation, laying the foundation for a potential global standard in AI governance, risk management, and transparency. A key pillar of the Act is data transparency, i.e., the ability to understand and audit training data, algorithmic processes and AI output. 

For example, the Act establishes the following obligations:

• During AI training (machine learning): Ensuring high-quality, unbiased, and representative training datasets.
• Inside the black box: Documenting and monitoring requirements related to algorithmic transparency, ensuring interpretability and traceability.
• AI output: Enabling human oversight, contestability, and post-market monitoring to mitigate risks and ensure accountability.

The AI Act also introduces transparency requirements to ensure that users are informed when interacting with an AI system. For example, AI systems generating synthetic content (such as deepfakes) must clearly label their outputs as artificially generated. Furthermore, providers of general-purpose AI models must meet specific disclosure obligations, including transparency regarding copyrighted content used for training models.

At the same time, questions remain about how the transparency obligations in the AI Act interact with other legal frameworks, including:

• Fundamental rights under the EU Charter
• Copyright protection for content used as training data in AI development
• Trade secret protection for training data and AI system operations
• Obligations under the General Data Protection Regulation (GDPR) regarding personal data protection
• Competition law restrictions on data access and potential abuse

Against this backdrop, this workshop—featuring leading experts, policymakers, and industry professionals—will explore how these requirements translate into real-world AI compliance strategies and what challenges remain in opening the black box while preserving innovation and competitiveness.

When & where: 1 October 2025. 

• in Stadshallen (Lund City Hall, Lund, Sweden: 9.00 to 15.30 CET
• online: 9.25 to 12.00 and 13.00 to 15.15  CET

Spoken language: English

Registration: To participate will be free of charge. 
You are welcome to register to participate on site in Lund or online at ai.lu.se

Confirmed Speakers

• Johan Axhamn, Business Law, Lund University
• Jacob Dexe, TBA
• Hajo Michael Holtz, Department of Law, Uppsala Universitet
• Janne Elvelid, Head of Public Policy Sweden and Finland, Facebook
• Behrang Kianzad, Business Law, Lund university
• Ida Koivisto, Law, University of Helsinki
• Stefan Larsson, Technology and Society, Lund University
• Jonas Ledendal, Business Law, Lund University
• Ana Nordberg, Law, Lund University
• Sebastian Schwemer, Law and Governance. Oslo Business School
• Kasia Söderlund, Technology and Society, Lund University
• Katja de Vries, Law, Uppsala University

Preliminar programme outline

09.00 Registration open
09.25 Online meeting open
09.30 Presentations & Panels
12.00 Lunch break
13.15 Presentations and Panels
15.15 Coffee & mingle

Registration 

To participate is free of charge. Welcome to sign up for participation on-site in Lund och online at  ai.lu.se/2025-10-01/registration. Online participants will get a link for access the meeting at the zoom platform a couple of days before the event.

Organisation

• Johan Axhamn, Senior lecturer, Department of Business Law, Lund University
• Behrang Kianzad, Researcher, Department of Business Law
• Eduardo Gill-Pedro, Associate senior lecturer, Department of Law, Lund University
• Jonas Ledendal, Senior lecturer, Department of Business Law, Lund University
• Stefan Larsson, Senior lecturer, Department of Technology and Society
• jonas [dot] wisbrant [at] control [dot] lth [dot] se (subject: AI_in_EU_LAW_26_sept) (Jonas Wisbrant), AI Lund

• "Fika-to-fika" means that we start the day, not to early, with a Swedish fika with coffee and refreshment and end the day not to late with an afternoon "fika" with the ambition to leave plenty of room in the programme for discussions and networking.