As a parting gift, in December the Hungarian Council presidency published its summary of Member States’ contributions to the policy questionnaire on the relationship between generative Artificial Intelligence and copyright and related rights.
The questionnaire, which was presented at the Copyright Working Party meeting of 3 July 2024 gave Member States the opportunity to provide their input on a number of issues relating to Artificial Intelligence and copyright as well as share national experiences. By publishing the summary of the findings, the Hungarian Council presidency is providing insights into the thoughts and priorities of Member States and giving an indication of the Council’s priorities in the next couple of years.
A clear “no” to new legislation
Generally speaking, there is a noticeable amount of consensus across Member States on some of the fundamental questions relating to AI and copyright in the EU.
Most importantly, the Member States agree that the current EU framework is overall sufficient to regulate the technology. While demands were raised to clarify the existing acquis and to provide more clarity on some aspects of the ruleset, the responses to the questionnaire make it clear that there is no interest in new legislation.
Member States follow the same line of thinking when it comes to the question of copyright protection for AI-generated works. In line with one of our latest blog posts, the Member States argue that purely autonomously generated outputs should not be eligible for copyright protection. When there is significant human input in the creative process, such works may still qualify for protection. Like us, Member States deem that the existing copyright framework is equipped to handle such cases and there is no need to provide any kind of new or additional copyright-related protection, including a new sui generis right.
It is very reassuring to see the Member States’ desire to clarify and work with the existing legislative framework instead of introducing new legislation. In particular in light of the recently adopted AI Act and the ongoing work of the AI office, expanding the legal framework at this stage would bear significant risks of creating legal uncertainties and overcomplicating an already complex legal landscape.
The elephant in the room
Unfortunately the Council’s aforementioned support of existing legislation falls short when it comes to the use of copyright exceptions for the training of AI systems, as the majority of Member States argue that the training of AI systems goes beyond the scope of the copyright exceptions under the DSM Directive. This directly contradicts the AI Act that was just adopted with the Council’s support last year, which states that the DSM’s copyright exceptions allow reproductions and extractions of works or other subject matter for the purpose of text and data mining (Recital 105).
However, most Member States seem to have faith that the AI Act will strengthen the bargaining position of rights holders vis-à-vis AI developers. In fact, Member States largely oppose to re-open the issue, instead suggesting to “wait and see” whether the AI Act can deliver the desired results.
Particular hope is placed on the AI Act’s transparency provisions (see our reporting on the code of practice consultations) as several Member States note that the opt-out mechanism in combination with the transparency requirements will be sufficient to encourage the creation of a licensing market to ensure compensation for authors whose works are used to train AI. Alternatively, a handful of Member States have issued their support for a remuneration scheme. However, no proposals were put forward regarding what shape such a system could take.
It is worrying to see that only a few months after its adoption, Member States are already ready to second-guess certain elements of the AI Act. However, it does not seem like this position will manifest itself in more concrete actions so long as the implementation of the transparency provisions happens in a clear and reliable manner.
Beyond these points of relatively broad alignment, the summary of contributions to the questionnaire also provides insights into some of the more niche positions of Member States on the issue of copyright and AI. Unfortunately, the summary does not attribute ideas to individual Member States so it is unclear where exactly these ideas originate. Some notable contributions include one Member State echoing a rightsholder position proposing a domaine publique payant for outputs generated exclusively by AI, and another supporting the use of blockchain technology to make AI-generated content distinguishable from human creation. As both contributions were only brought forward by a single Member State, it is unlikely that anything actionable will arise from these ideas any time soon.
