Remarks on the protectability of generative AI outputs
By Alexander Thesleff[*]Doctoral Researcher, Faculty of Law, University of Helsinki.
“Can works created using AI be protected?” The research findings from my ongoing dissertation work from the Nordic/Finnish perspective point towards a quite succinct “yes”. This stems already from the wording of this years’ symposiums’ debate question: “works” will hint towards outputs that are sufficiently original, and “using” presupposes a (natural) person interacting with the technology. But as there exists no authoritative case law at the EU-level or in the Nordics on the matter, the question is still highly topical and far from settled. Lately the U.S. approach has been very negative towards generative AI’s (GenAI) creations’ copyrightability, and such a stance is arguably not fully misguided either.[1]USCO review board refusal in Théâtre D’opéra Spatial -registration (final decision of September 5, 2023), https://www.copyright.gov/rulings-filings/review-board/docs/Theatre-Dopera-Spatial.pdf; USCO letter of February 21, 2023 to Kristina Kashtanova re cancellation of registration of Zarya of the Dawn, https://www.copyright.gov/docs/zarya-of-the-dawn.pdf; USCO: Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence March 16 2023 16190 Federal Register, Vol. 88, No. 51 Rules And Regulations 37 Cfr Part 202. https://www.copyright.gov/ai/ai_policy_guidance.pdf.
Firstly, to touch upon the division between “AI-generated” and “AI-assisted” creation: It can be a useful distinction to make, but, I would argue, only on the very high level of whether or not there exists a user of an AI-system in the first place. When looking at e.g. the contemporary large language models (LLM) and diffusion models utilized in most of the AI-facilitated text and image creation, there is always somebody, who has a need or a vision for a creative product, who – driven by that need and other varying case-by-case parameters – engages with the technology, and reviews critically what the system generates. At least in a broad sense we have causation[2]For a comprehensive discussion of causation in copyright law, see Rognstad, O-A: Creations Caused by Humans (or Robots)? Artificial Intelligence and Causation Requirements for Copyright Protection in EU Law in Artificial Intelligence and the Media: reconsidering rights and responsibilities (Edward Elgar 2022), pp. 172–191. See also Axhamn J: Copyright and Artificial Intelligence – with a focus on the area of music, in Festskrift til Jørgen Blomqvist (eds. Rosenmeier et al.), Ex Tuto Publishing 2021, p. 33–86. of the output reaching back to the user who has chosen the specific AI tool, prompted the AI in a specific manner[3]It can be said that a decisive user outlined here possesses the prerequisite authorial intent as elaborated by Hugenholtz P B – Quintais J P: Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output? IIC (2021) 52:1190–1216, p. 1199–1206. and exercised human discretion that ultimately dictates which creation is chosen (or refined).
That being said, it is not impossible to imagine completely independent AI-systems that are self-actuating, self-monitoring and even self-publishing. In these cases we should be talking about authorless creations that fall outside the anthropocentric premise of copyright.
