A statutory model of compensation for the use of works in artificial intelligence training as a proposal for systemic strengthening of copyright in the era of machine learning
Abstract
The dynamic development of generative artificial intelligence has revealed the structural inadequacy of traditional copyright mechanisms for the mass, automated processing of data used in model training processes. Classic licensing systems, based on identifiable and individual access to a work, are unable to function in an environment where a single algorithm processes millions of protected content in a way that cannot be reproduced using existing regulatory tools. The article analyzes the legal and economic rationale behind the proposal to introduce a statutory compensation model for the use of works in artificial intelligence training, inspired by the European system of reprographic fees. The study attempts to demonstrate that this model can be an effective tool for balancing the interests of creators and technology companies, while strengthening the stability of the copyright system. It also points to the need to supplement future regulations with data transparency obligations and appropriate institutional instruments to enable the efficient redistribution of remuneration. The results of the analysis lead to the conclusion that statutory compensation can become the foundation for the sustainable development of the creative and technological ecosystem in an era of growing AI autonomy.
References
- Floridi, L., & Chiriatti, M. (2020). GPT-3: Its nature, scope, limits, and consequences. Minds and Machines, 30, 681–694.
- Geiger, C. (2021). The missing goal-scorers in the artificial intelligence team: Of big data, the fundamental right to research and the failed text and data mining limitations in the CDSM Directive. In Intellectual property and sports: Essays in honor of P. Bernt Hugenholtz (pp. 383–394). Kluwer Law International.
- Geiger, C., & Iaia, V. (2024). The forgotten creator: Towards a statutory remuneration right for machine learning of generative AI. Computer Law & Security Review, 52, 1–9.
- Geiger, C., Frosio, G., & Bulayenko, O. (2018). Crafting a text and data mining exception for machine learning and big data. In Intellectual property and digital trade (pp. 95–112).
- Geiger, C., Frosio, G., & Bulayenko, O. (2018). Text and data mining in the proposed copyright reform. International Review of Intellectual Property and Competition Law, 49, 814–844.
- Hugenholtz, P. B. (2012). The story of the tape recorder and the history of copyright levies. In Copyright and the challenge of the new. Wolters Kluwer.
- Lemley, M. (2024). How generative AI turns copyright upside down. Columbia Science & Technology Law Review, 25(2), 190–212.
- Lucchi, N. (2023). ChatGPT: A case study on copyright challenges for generative AI systems. European Journal of Risk Regulation, 15(3), 602–624.
- Margoni, T., & Kretschmer, M. (2022). A deeper look into EU TDM exceptions. GRUR International, 71, 685–701.
- Netanel, N. (2003). Impose a noncommercial use levy to allow free peer-to-peer sharing. Harvard Journal of Law & Technology, 17(1), 2–83.
- Oksanen, V., & Välimäki, M. (2005). Copyright levies as an alternative compensation method. Review of Economic Research on Copyright Issues, 2(2), 25–39.
- Senftleben, M., Margoni, T., et al. (2022). Ensuring visibility and accessibility of European creative content. JIPITEC, 13, 67–86.
- Truyens, M., & Van Eecke, P. (2014). Legal aspects of text mining. Computer Law & Security Review, 30, 153–170.
- Wang, J. T., Deng, Z., et al. (2024). An economic solution to copyright challenges of generative AI. arXiv.