Introduction
The rapid development of artificial intelligence (AI) technology has significantly impacted the production and dissemination of creative content. As AI-generated works proliferate across the European Union (EU), legal questions arise regarding the ownership, protection, and distribution of these works under copyright law. Directive (EU) 2019/790, commonly known as the Directive on Copyright in the Digital Single Market (DSM Directive), seeks to modernize the EU’s copyright framework in light of digital and technological advancements, including AI. While the Directive does not provide explicit provisions for AI-generated content, its principles offer guidance on how such content might be treated within the existing legal framework.
This article examines the legal landscape surrounding AI-generated content in the EU under the DSM Directive and discusses the challenges and ambiguities posed by the increasing role of AI in creative industries.
The Scope of Copyright Protection in the EU
Under EU law, copyright protects original works of authorship. For a work to qualify for protection, it must be original, meaning that it must reflect the author’s own intellectual creation. The notion of originality, rooted in the jurisprudence of the Court of Justice of the European Union (CJEU), requires that a work express the personality and free choices of a human author.
The DSM Directive reaffirms this traditional requirement of human authorship. Recital 5 of the Directive emphasizes the need to protect the legitimate interests of authors and rightsholders in the digital age. However, AI-generated works challenge the fundamental premise of human authorship since these works can be created autonomously by machines without direct human intervention. As such, the question arises: can AI-generated content qualify for copyright protection under the DSM Directive?
AI-Generated Content and the Concept of Authorship
The DSM Directive does not define “authorship” in relation to AI-generated content, nor does it introduce specific provisions governing AI as a creator. This omission reflects the broader uncertainty in international copyright law concerning AI-generated works. Given that copyright protection in the EU is traditionally predicated on the idea of a human author, works produced solely by AI systems are unlikely to qualify for protection under existing copyright laws.
The key issue is whether an AI-generated work can meet the requirement of “originality” if no human involvement is present. The CJEU has made it clear in cases like Infopaq International A/S v Danske Dagblades Forening (C-5/08) and Eva-Maria Painer v Standard VerlagsGmbH (C-145/10) that copyright protection requires that copyrighted works “contain elements which are the expression of the intellectual creation of the author of the work”, and that “an intellectual creation is an author’s own if it reflects the author’s personality”, criteria which cannot be satisfied by mere mechanical or automated processes. Consequently, fully autonomous AI-generated content would not meet the originality criterion necessary for copyright protection.
A similar trend can be noted in the jurisprudence of U.S. courts which have consistently held that content generated solely by artificial intelligence (AI) without meaningful human involvement is not eligible for copyright protection. This principle was notably affirmed in the 2023 case Thaler v. Perlmutter, where the U.S. District Court for the District of Columbia ruled that works lacking human authorship do not qualify for copyright.
Similarly, in 2023, the U.S. Copyright Office denied registration for “Théâtre D’opéra Spatial,” an image created using the AI tool Midjourney. The Office concluded that AI-generated elements dominated the work and human input was minimal, rendering it ineligible for copyright protection.
All these decisions underscore the legal requirement for human creativity in works seeking copyright protection, a criterion which excludes purely AI-generated content from such rights.
Human Input in the Creative Process
Despite the absence of explicit rules for AI-generated works, the DSM Directive offers insights into how AI-assisted works, where there is some degree of human input, may be treated. For instance, if a human author uses AI tools to generate content but exercises creative control over the process, the resulting work could potentially qualify for copyright protection. This is contingent on the extent to which the human author contributes to the originality of the work.
In this context, the AI system can be seen as a tool, much like a camera or a computer, which facilitates the human author’s creative expression. As long as the human author makes creative decisions that contribute to the originality of the work, the work may be eligible for copyright protection under the existing system.
Ownership of AI-Generated Content
Another critical issue is determining the ownership of AI-generated content. Under the DSM Directive and EU copyright law, the author is the initial owner of the copyright, and any rights associated with the work (such as reproduction, distribution, and public performance) vest in the author. In the case of AI-generated content, the question arises: who owns the rights to a work created by an AI system?
The DSM Directive does not address this issue directly, leaving room for interpretation. In scenarios where AI is used as a tool by a human creator, the human operator who uses the AI and provides the necessary creative input would likely be regarded as the copyright owner. However, when the AI system operates independently of human intervention, it is unclear who, if anyone, can claim ownership of the resulting content. Current EU law does not provide for AI systems to hold rights, nor does it recognize non-human entities as authors.
To address this gap, some have suggested that the developer or owner of the AI system could be entitled to ownership of the works produced by the system. However, this approach raises further questions about whether the mere provision or development of AI technology is sufficient to confer copyright ownership.
Text and Data Mining (TDM) and AI
One area where the DSM Directive explicitly intersects with AI is in the provisions governing text and data mining (TDM) (i.e., the use of copyright-protected materials by generative AI tools to train their algorithms).
Articles 3 and 4 of the DSM Directive introduce exceptions to copyright for TDM, which is a technique often employed by AI systems to analyse vast amounts of data. TDM plays a crucial role in AI development, as it enables AI systems to “learn” by analysing large datasets. The DSM Directive’s TDM exceptions thus facilitate AI innovation by allowing researchers and developers to use copyrighted works as training data for AI systems (subject to compliance with the Directive’s requirements). However, the Directive does not extend TDM exceptions to all users.
TDM is considered lawful in the EU under the following conditions:
1. For scientific research by specific institutions: An exception exists for the rights outlined in Articles 5(a) and 7(1) of Directive 96/9/EC, Article 2 of Directive 2001/29/EC, and Article 15(1) of the DSM Directive. This applies to reproductions and extractions made by research organizations and cultural heritage institutions for the purpose of scientific research, provided they have lawful access to the works and other subject matter intended to be data mined (as per Article 3 of the DSM Directive).
2. For general use by any institution: Another exception applies to the rights granted under Articles 5(a) and 7(1) of Directive 96/9/EC, Article 2 of Directive 2001/29/EC, Articles 4(1)(a) and (b) of Directive 2009/24/EC, and Article 15(1) of the DSM Directive. This exception permits reproductions and extractions from lawfully accessible works by any institution, as long as the works being mined were accessed lawfully and the copyright owner has not expressly prohibited their use for data mining purposes (as per Article 4 of the DSM Directive).
Both exceptions are subject to Article 5(5) of Directive 2001/29/EC, which states that those limitations or exceptions will only be applied for those “certain special cases which do not conflict with a normal exploitation of the work or other subject-matter and do not unreasonably prejudice the legitimate interests of the rightsholder”.
The exception provided in Article 3 allows research organizations and cultural heritage institutions to mine copyrighted works without the rightsholder’s authorization under certain conditions. Commercial entities must either obtain licences or rely on the limited exceptions provided in Article 4.
However, the distinction between these two exceptions is not as clear-cut as it may seem: a German court decision last October (LG Hamburg 310 O 227/23), the first published decision by an EU national court interpreting the scientific research and general use TDM exceptions, demonstrates this. The case concerned a non-profit organization that relied on the scientific research exception (Article 3) to create a large dataset of image-text pairs. Although this dataset was created under the guise of non-commercial research, it was made publicly available for commercial AI training. The court interpreted the scope of the scientific research exception in such a way that the described result is generally legal (and would not be allowed only if a private for-profit company controlled the research institution and had granted privileged access to the results of the research). While the decision has been welcomed by AI developers, it raises alarm bells for copyright holders: the risk of creating a loophole where companies claim non-commercial purposes to justify large-scale data mining, which authors can’t object to, but the resulting datasets can be exploited commercially, undermines copyright protection. It is clear that further case law will be needed before the rules are clearly settled.
***
We remain at your disposal to provide more information and/or related legal assistance connected to this topic, via our intellectual property global practice which comprises lawyers across all our offices.
No information contained in this article should be considered or interpreted in any manner as legal advice and/or the provision of legal services. This article has been prepared for the purposes of general information only. PETERKA & PARTNERS does not accept any responsibility for any omission and/or action undertaken by you and/or by any third party on the basis of the information contained herein
