Rainer Mühlhoff, Hannah Ruschemeier

Journal of AI Law and Regulation, Volume 1 (2024), Issue 1, Page 24 - 39

This article proposes the concept of purpose limitation for AI models as an approach to effectively regulate AI. Unregulated (secondary) use of specific models creates immense individual and societal risks, including discrimination against individuals or groups, infringement of fundamental rights, or distortion of democracy through misinformation. We argue that possession of trained models, which in many cases consist of anonymous data (even if the training data contains personal data), is at the core of an increasing asymmetry of informational power between data companies and society. Combining ethical and legal aspects in our interdisciplinary approach, we identify the trained model, rather than the training data, as the object of regulatory intervention. This altered focus adds to existing data protection laws and the proposed Artificial Intelligence Act. These are inefficient in preventing the misuse of trained models due to their focus on the procedural aspects of personal data or training data. Drawing on the concept of risk prevention law and the principle of proportionality, we argue that the potential use of trained models by powerful actors in ways that are damaging to society warrants preventive regulatory interventions. Thus, we seek to balance the asymmetry of power by enabling democratic control over where and how predictive and generative AI capabilities may be used and reused. Keywords: EU AI Act; GDPR; purpose limitation; regulating models; data power

Mark D. Cole

Journal of AI Law and Regulation, Volume 1 (2024), Issue 1, Page 126 - 142

Nelly Stratieva

Journal of AI Law and Regulation, Volume 1 (2024), Issue 1, Page 102 - 109

An agile approach to the EU AI Act and its surrounding ecosystem of hard and soft law can promote more dynamic AI governance in the EU. Agile has powered the rise of AI; it originates from the field of software development and is a style of working that embraces change and brings fast, tangible value to stakeholders through iterative product development. In this opinion I highlight two challenges that the AI Act faces and discuss how an agile mindset can be applied to mitigate their negative effects. I advocate that the AI Act ecosystem can, and should, be improved in iterations based on real-life testing, informed by the EU’s own Better Regulation toolbox and international best practices in innovative, agile regulation. Keywords: EU AI Act; agile; better regulation

Boniface de Champris

Journal of AI Law and Regulation, Volume 1 (2024), Issue 1, Page 113 - 117

New developments in artificial intelligence (AI) have triggered heated legal and political discussions globally, including in the field of copyright. With significant legal cases pending in the United States, the debate in the European Union (EU) has mainly been political. The recent political agreement on the EU’s AI Act will likely reverse the delicate balance reached by the EU Copyright Directive and could endanger innovation. Keywords: copyright; EU AI Act; generative AI; TDM

Tatjana Evas

Journal of AI Law and Regulation, Volume 1 (2024), Issue 1, Page 98 - 101

The European Union’s Artificial Intelligence Act (AI Act)1 is one of the most widely discussed and anticipated pieces of upcoming legislations, capturing global attention with its approach to governing uses of AI technologies. Following intense preparatory work,2 and almost three years of complex political negotiations, culminating in political agreement on 8 December 2023, the AI Act is set to shape global regulatory space on how we innovate, manage, and perceive AI technologies. This opinion, drawing upon first-hand experience from the AI Act’s conceptualisation, legal drafting, and negotiations, offers a reflection on key decisions and the overall rationale behind this innovative and pioneering regulation. Keywords: EU AI Act; trustworthy AI; risk-based approach; horizontal framework; product safety