This post is the second instalment of an analysis of a recent report, a part of the reCreating Europe project, on the application of EU copyright and related rights law to outputs generated by or with the assistance of artificial intelligence (AI) systems, tools or techniques (AI outputs), with a focus on outputs in the musical domain. The report examines the question: How can and should EU copyright and related rights law protect AI musical outputs? The first part summarised the conclusions of our research on how EU copyright rules to apply to AI music outputs. This second part presents our conclusions regarding EU rules on related rights and outlines policy recommendations for EU legislators in this field.
In the music field, the practical significance of the uncertainties regarding copyright protection of AI outputs is mitigated to some extent as commercial actors can rely on related rights protection.
Performances and performers
The relationship between works and performances is a conventional minimum for granting related rights to performers. As EU law does not make protection of performances conditional on the performance of works, Member States are free to grant related rights for the performance of subject matter that is not protected by copyright. There is no uniform approach among the Member States to this question. Presently, the performance of non-copyright protected AI-generated outputs is protected in some Member States but not in others.
Performance by the Philharmonic Orchestra of Luxembourg of a musical piece by AIVA (video)
Artistic performance is commonly a technologically neutral act. While performers are free to choose any means, performance requires some active involvement on the side of the artist to be granted related rights protection. A mere act of activating AI-enabled speech or sound generation (e.g., by pushing the button “generate”) without any further action is not a performance in a sense intended by the international treaties. The technological neutrality of related rights protection also means that they are granted to performers using AI systems as tools for enhancing, altering or transforming their performances.
Performers’ original improvisation with AI tools and techniques, like with traditional instruments, is protected as a copyright work if it meets the originality requirements and fulfils the condition of fixation, where it is required by national law.
Phonograms and phonogram producers
Related rights protection of phonograms is independent of that of copyright or performances. As such, phonogram producers benefit from legal protection regardless of whether the underlying sound was created by a human or AI-generated. There is no threshold for protection other than that for the subject matter to qualify as a phonogram, i.e., it must be a fixation of sound. The fixation requirement implies that AI-enabled continuous unfixed music generation and streaming projects are not phonograms and hence are not covered by rights awarded to phonogram producers. Our research did not identify significant legal issues with the qualification of recordings of music produced with the aid of AI systems as phonograms or with the corresponding attribution of rights.
Broadcasts and broadcasters
Protection of broadcasts is independent of the subsistence of copyright and related rights of performers and phonogram producers to the broadcasted content. The use of AI systems in music production does not seem to have any significant impact on the legal qualification of broadcasts or the protection of broadcasters. In the Member States where broadcasting protection extends to webcasting, such protection could apply to AI-enabled continuous unfixed music generation and streaming projects. For an example of such projects, see “EōN” of Jean-Michel Jarre and “Reflection” of Brian Eno.
As a general remark, the availability of related rights protection for non-copyright protected AI-generated outputs should, from a normative and policy standpoint, be taken into account when assessing the desirability of proposals for new modes of protection of AI outputs.
None of the consulted stakeholders and experts relies on or considers the existing national regimes of computer-generated works to be of particular importance for AI music production. In this regard, economic actors appear to rely on the familiar copyright and related rights protection rather than on the specific regime for computer-generated works. Our research found no evidence that the establishment of any AI music services studied in a particular jurisdiction was motivated by the existence of legal protection for computer-generated works. This finding also suggests that policy proposals for a legislative change based on this legal model should be considered with particular caution.
On the basis of the analysis, out report advances the following recommendations.
- There is no clear case for a legislative action at the level of substantive rules in the EU copyright acquis in the short term as regards AI outputs. Existing proposals for new rights and forms of protection for AI outputs generally lack clear and convincing theoretical and economic justification. In most cases, these proposals fail to adequately consider existing protection for AI outputs under copyright law and, where such protection is lacking, under related rights or (in limited cases) specific regimes for protection of computer-generated works. Considering this, we recommend that no new protection regimes for AI outputs are introduced absent clear and compelling evidence that justifies a change to the status quo.
- With regard to the protection of performers, given the increasing frequency and scale of performances of AI music outputs, it is recommended in the short term to carry out a mapping analysis of whether and how Member States’ laws grant of related rights to performers is conditional on the performance of “works”. Taking into account the uncertainties of qualifying AI outputs as “works”, it is recommended in the medium term to consider EU harmonisation of the requirement for granting related rights to performers independently from the copyright status of the content performed.
- The flexibility of the requirements for copyright protection and authorship at the EU level provides private parties concerned with some interpretative space. Private actors are experimenting with different contractual arrangements for achieving desired legal certainty and rights attribution. Open disputes between the parties on the subsistence of copyright protection and/or authorship are rare. As such, and absent concrete evidence to the contrary (particularly of economic nature), it is recommended that the development of artistic, business, and contractual practices is closely monitored and subject to further study. Future work in this respect at the international or EU level could include stakeholder dialogues and co-regulatory approaches with a view to identifying and developing best practices and model clauses to guide AI service providers in this area.
- In the medium term, it is recommended to scrutinise the presumption of authorship and ownership in Art. 5 Enforcement Directive. The assessment of this provision should focus on the areas: (i) where a declared absence of authors could spare economic actors from some copyright-related costs (e.g., royalty payments to authors); and (ii) where the declared presence of authors could create copyright-related revenues (e.g., based on copyright protection of the AI outputs). Users of AI systems should retain the right to claim authorship over AI outputs that qualify as works as a result of their contribution, as well as have recourse to legally effective means to disclaim authorship of AI outputs or parts thereof. Further research should focus on the legal mechanism(s) that could achieve these goals in the context of a revision presumption of authorship and ownership and/or the right to object to false attribution.
- The proposed four-step test for assessing copyright protection of output created with AI as a “work” provides a solid analytical basis for reducing the legal uncertainty in this area and a strong descriptive capacity in EU copyright law for the attribution of copyright protection to human authors. Although further research is needed to develop a corresponding normative argument, we have identified the notion of “human cause” as a promising avenue in that direction. Based on this normative consideration and to address existing uncertainties, it is recommended that legislative and/or judicial authorities authoritatively affirm the normative anthropocentric conception of EU copyright protection, thus guiding the application of copyright rules to AI outputs.
This research is part of the reCreating Europe project, which has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 870626.