Access to uncontaminated data for AI training
International Journal Of Law And Information Technology · 12.01.2026 · 34
DOI: 10.1093/ijlit/eaag012 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal Of Law And Information Technology, online erschienen am 19.08.2026. Verfasst von Maurice Chiodo und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Large language models and other generative AIs are now contaminating the online data environment with vast quantities of AI-generated—‘synthetic’—content, which is difficult to filter out. This poses a problem for those training future AIs due to the risk of model collapse, whereby an AI trained on synthetic content produces increasingly useless or undesirable output. This can entrench existing players in their dominant positions over newcomers by virtue of the ‘uncontaminated’ (ie, synthetic free) data they collected prior to the proliferation of generative AI in 2022. In this article, we address questions about access to uncontaminated data and other essential inputs for AI training, by considering existing legal regimes such as AI regulation, data (access) governance, (European Union and domestic) competition law, and gatekeeper rules for digital markets. We show that there is no tailor-made solution in existing laws. However, various elements reinforce the idea that access to an essential resource, which is increasingly scarce, and at significant risk of ‘extinction’, is something lawmakers should urgently address. We conclude with considerations on how such data access could be afforded in a fair and equitable manner—while aiming to minimize the risks of harmful uses resulting from such access.
Zitieren
APA
Chiodo, M., Ruse-Khan, H. G., Markschies, L., Müller, D., Podszun, R., & Zech, H. (2026). Access to uncontaminated data for AI training. International Journal Of Law And Information Technology, 34. https://doi.org/10.1093/ijlit/eaag012
Vancouver
Chiodo M, Ruse-Khan HG, Markschies L, Müller D, Podszun R, Zech H. Access to uncontaminated data for AI training. International Journal Of Law And Information Technology. 2026;34. doi:10.1093/ijlit/eaag012
BibTeX
@article{Chiodo2026,
author = {Chiodo, Maurice and Ruse-Khan, Henning Grosse and Markschies, Lisa and Müller, Dennis and Podszun, Rupprecht and Zech, Herbert},
title = {Access to uncontaminated data for AI training},
journal = {International Journal Of Law And Information Technology},
year = {2026},
volume = {34},
doi = {10.1093/ijlit/eaag012}
}
EUVIMED