Joint Statement on Generative Artifical Intelligence and the EU AI Act

The undersigned organisations represent the collective voice of hundreds of thousands of writers, performers, composers, songwriters, screen directors, screenwriters, visual artists, journalists, translators and other creative workers whose human artistry lies at the core of the creativity that our societies cherish and enjoy on a daily basis.

As the EU AI Act is expected to soon enter into force, we welcome the adoption of this regulation, which is a first step in the right direction and the first attempt by a major regulator to harness the unfettered use of AI, including general purpose AI models. Despite the tech industry’s resistance to future regulation, this technology and its use have long-ranging implications for individuals, society, creators, and cultural diversity. It is of utmost importance to make sure it can develop in a way that respects fundamental principles and as a tool to enhance, not replace, human creativity.

In particular, we welcome the introduction in the EU AI Act of specific rules for providers of general-purpose AI models and notably the requirement for them to comply with EU copyright law and publish sufficiently detailed information about the data used. We also strongly support the strengthening of transparency obligations around deep fakes and stress the importance of developing technical tools that may reliably and accurately differentiate authentic content from AI-generated, or manipulated, content. We thank the European Parliament for championing those provisions during the inter-institutional negotiations.

The AI Act still needs to be implemented in an effective way in order to preserve fundamental rights, safeguard transparency, and enable authors and performers to exercise their rights. We will remain vigilant about this implementation, both at EU (codes of practice, template of the AI office, etc.) and national level. Without sufficient transparency, it will be impossible for our members to avail themselves of the protection provided by the acquis, both related to copyright and data protection.

However, even with a proper implementation of the AI Act, it will only serve as a temporary fix for a much larger problem unless legal ambiguities are addressed and transparency, informed consent and remuneration are firmly enshrined in the existing legal framework.

The AI Act new section aimed at providers of general-purpose AI, including providers of large generative AI models, refers to the implementation of the text and data mining exception for commercial purposes (Article 4 of EU Directive 2019/790), but does not effectively extend the scope of this exception. Under EU law, a copyright exception should always be restrictively interpreted and there is no decision from the Court of Justice on this topic. Extending the scope of the TDM exception to generative AI would place creators before an unacceptable and undemocratic “fait accompli” letting the tech industry off the hook for having used creators’ work with generative AI without their permission and in complete opacity for a significant period of time.

Directive 2019/790, and the TDM exception in its Article 4, does not mention or define “Artificial Intelligence” and “Generative AI” and was not conceived with large scale generative AI models in mind. In the absence of any relevant court decision, expanding this exception to cover the systematic and extensive use of creators’ protected works and performances, with the intention of generating synthetic content for commercial use that closely resembles, and therefore directly competes with their work, is not only unfair but also legally questionable. It is hard to conceive how this exception, which is poorly defined as to how it may be exercised in relation to generative AI, would satisfy the three-step test, a fundamental safeguard intended to strike a fair balance between rightholders and content users by limiting copyright and neighbouring right exceptions to certain special cases that do not conflict with the normal exploitation of the works or other subject matter and do not unreasonably prejudice the legitimate interests of rightholders. Copyright exceptions are generally conceived to ensure a fair balance between intellectual property and other fundamental rights or public policies (such as freedom of expression or education) to the benefit of citizens or public institutions. This broad interpretation of the TDM exception does not meet any public policy objective and appears to retrospectively justify the massive scraping of our members’ works and performances, to the sole benefit of AI companies.

We would also like to stress that, while AI models scrape creators’ works at an unprecedented scale, any effective reservation of rights for TDM should already have led to licenses between all the original rightholders in the works concerned and generative AI businesses, including after the transfer of the creators’ exclusive rights to a third party. Creators should have received information about the revenues that the scraping of their works contribute to generate, in order to determine and enforce appropriate licensing mechanisms. This has not however been the case. The truth is that their works and personal data have been used on a massive scale without any possibility for them to reserve their rights and be remunerated in return. Authors and performers as well as their contractual counterparts should not bear the costs and be at the mercy of specific opt out mechanisms built by the tech industry through illusory self-regulation.

In addition, it is essential to stress that authors and performers generally assign or transfer their IP rights to their contractual counterparts, based on treaties, conventions and other normative instruments that were in no way designed to cover AI-related uses. Using works and performances in the context of AI models is radically different from traditional forms of exploitation. In this context, as the integrity of their work and their personal reputation may well be jeopardized, creators should retain the ability to consent or refuse such usage of their work. As such, they must be involved in the exercise of their rights reservation, the design of the technical protocols used to this end – as well as any policy discussions regarding generative AI.

As a new EU policy cycle is about to start, we urge EU decision makers to engage in a comprehensive and democratic debate leading to a clear legal framework preserving the rights and the integrity of the works of creators, addressing the numerous open issues linked to the TDM exception today and clarifying the terms of its possible extension to generative AI. EU policy makers must design an ambitious strategy for a truly human-centric generative AI that ensures that creators can exercise informed consent, be fairly remunerated for the use of their works through generative AI and receive financial compensation for all the current and future AI-produced contents that their creative work has contributed to generate.