Over five years have passed since the adoption of the EU Directive on Copyright and Related Rights in the Digital Single Market (‘DSM Directive’). As its title suggests, by adopting this directive, the EU legislature intended to tackle and realize the digital dimension of the EU single market from a copyright and related rights perspective. Yet, looking at the resulting national transpositions, it is highly doubtful that such an objective might be considered achieved. Meanwhile, the Court of Justice of the European Union (‘CJEU’) has been already referred questions on the interpretation of relevant provisions of the directive, in light of doubts regarding the compatibility between these and national transpositions thereof (ONB and Others, C-575/23 (Belgium); Meta Platforms Ireland, C-797/23 (Italy); Streamz and Others, C-663/24 (Belgium)).

The complexity of the situation is further increased by the adoption, after the DSM Directive, of horizontal legislation – including but not limited to the Digital Services Act (‘DSA’) – that, while not copyright-specific, also serve to construe relevant provisions of the directive and, insofar as the DSA in particular is concerned, be in a lex generalis to lex specialis relationship with, e.g., a key provision of directive like its Article 17. 

By focusing in particular on Articles 15 and 17 of the DSM Directive, as transposed into national law, this contribution reflects critically on the current state of EU copyright and considers what the future might hold for this area of EU law. It ultimately concludes that the approach to copyright harmonization is in need of serious rethinking, and so are the goals and the legislative instruments needed to in this regard.

Like all sectors, the movie industry has been both affected by and exploring potential uses of generative Artificial Intelligence ('AI'). On the one hand, movie studios have detected and begun to add warnings against unlicensed third-party uses of their content, including for AI training,1 and have taken enforcement initiatives through court action. On the other hand, the use of AI within and by the industry itself has been growing. Regarding the latter, some have emphasised the opportunities presented by the implementation of AI, including by advancing claims that AI tools can offer a `purer' form of expression. Others have instead warned against the potential displacement of industry workers, including workers employed in technical roles and younger and emerging actors. Against the background illustrated above, this study maps and critically evaluates relevant issues facing the development, deployment, and use of AI models from a movie industry perspective. The legal analysis is conducted having regard to EU and UK copyright law and is divided into three parts: • Input/AI training: By considering relevant legal restrictions applicable to the training of AI models on protected audiovisual content, the border between lawful unlicensed uses and restricted uses is drawn; • Protectability of AI-generated outputs: Turning to the output generation phase, the protectability of such outputs is considered next, by focusing in particular on the requirements of authorship and originality under EU and UK copyright law; • Legal risks and potential liability stemming from the use of third-party AI models for output generation: Still having regard to the output generation phase, relevant legal issues that might arise having regard to the use of AI models that `regurgitate' third-party training data at output generation are considered, alongside the question of style protection under copyright. The main conclusions are as follows: • Input/AI training: Insofar as model training on third-party protected content is concerned, there are no exceptions under EU/UK law that fully cover the entirety of these processes. As a result, lacking legislative reform, the establishment of a licensing framework appears unavoidable for such activities to be deemed lawful; • Protectability of AI-generated outputs: The deployment of AI across various phases of the creative process does not render the resulting content unprotectable, provided that human involvement and control remain significant throughout, with the result that AI is relied upon as a tool that aids – rather than replaces – the creativity of industry workers. • Legal risks and potential liability stemming from the use of third-party AI models for output generation: The use of AI models that generate infringing outputs, such as by regurgitating input data or merely imitating style, may trigger the application of exclusive rights under copyright and related rights. The resulting liability may vest with the user of such models, as well as the model developer/provider. The latter aspect means that terms that exclude any such liability may ultimately be found to be unenforceable against users and ineffective against rightholders.

• In autumn 2024, the Court of Justice of the European Union (CJEU) issued its judgment in Kwantum, C-227/23. Endorsing the Opinion of Advocate General (AG) Szpunar in the same case, the Court held that EU Member States do not enjoy any freedom to set their own reciprocity clauses under Article 2(7) of the Berne Convention: in this field, the competence vests with the EU and not (or rather: no longer) individual Member States. Like the AG, the decision further confirms that any work protected by copyright—including works of applied art—must only satisfy a requirement of originality, this also being a harmonized notion under EU law.

• The outcome of the referral from the Dutch Supreme Court was both unavoidable and unsurprising having regard to settled CJEU case law in the field of copyright, well beyond the immediate ‘precedent’ set by RAAP, C-265/19. On a broader level, it shows how a ‘Union copyright’ does not only exist but also extends well beyond the formal legislative framework.

hould the EU unify the copyright laws of its Member States and introduce, over 30 years after the start of the harmonization process, a unitary copyright title? The short answer is: yes. And that will be unavoidable too.Therefore, the real question is another: how to go about doing that? In order to answer this, it is necessary to start from the beginning and retrace the steps of a harmonization process that has led to a framework, which is an understatement to call ‘complex’.

In its 2009 decision in Infopaq (C–5/08), the Court of Justice of the European Union (CJEU) achieved a landmark result: the de facto, horizontal harmonization of the originality requirement. After that, nothing could stay the same. After providing an overview of the harmonization process in the copyright field over the past thirty years or so and, with that, the environment in which Infopaq came to be, this chapter considers Infopaq and the expansive effect of subsequent case law on other copyright subsistence requirements. The analysis also notes how the eventual outcome of Cofemel (C–683/17), insofar as works of applied art are concerned, is perfectly in line with such jurisprudence. The chapter further considers the legal and institutional difficulties that such a string of CJEU decisions has given rise to, and is yet to resolve, before concluding that further questions are likely to be posed to the CJEU in the not too distant future. In other words: the construction of EU copyright is far from over.