“[The author] shall […] provide the publisher with the exclusive right of reproduction and distribution (publishing right).” – This provision, set out in § 8 of the German Publishers’ Rights Act (VerlagsG), outlines the common practice for publishing an article in a journal, not only in Germany, but in many other countries too. What appears to mark the end of the author’s control over their article, is arguably just the beginning.

Scope of the granted publishing right

Authors grant publishers exclusive rights due to a division of tasks: Once the contribution has been finalised, the publisher assumes full exploitation of it (including the publication and distribution of the work). In the theoretical scenario of extensive use of the contribution and appropriate remuneration of the author, no further granting or return of rights would be necessary. In theory, every potential reader could access the work, and all possible revenue could be generated by publication. The author, the publisher and the readers would be satisfied.

In practice, however, things look different. For one, only a few publishing companies have significant market power and the ability to impose their general terms and conditions on authors. The readers’ willingness to pay to access the publications is likely far from being adequately addressed. Furthermore, a publication’s reach is inherently limited by visibility constraints, high costs (subscription and transaction), language barriers, and disciplinary barriers. Finally, both the scientific community and the general public have a legitimate interest in accessing works free of charge in order to build on them, particularly if the works were funded by public money.

Consequently, the right granted to the publisher is increasingly considered as the publisher’s “last” opportunity to exploit the work. However, due to limitations of exclusive rights either by agreement or by law, the author can exploit their own work further. This possibility becomes even more relevant when new use cases arise that were not covered by the original contract, such as the internet at the end of the last century and AI training today.

This analysis, though, focuses on selected aspects and does not discuss general rights of revocation (e.g., §§ 41, 42 of the German Copyright Act – UrhG) or the possibility of secondary use after ten years (§ 40a UrhG) or after one print edition (§ 5 VerlG).

Right to secondary use

Some journals offer authors a secondary right to use their work after an embargo period (often one year). If they don’t, Copyright law comes into play for collections, but not for any monographs.

While European law establishes a framework for Copyright law (e.g., the InfoSoc Directive and the DSM Directive), it is up to the Member States to specify conditions such as a mandatory right to secondary use. Under German law, specifically § 38 UrhG, the legislator balances the interests of publishers of collections and authors by granting the publisher an exclusive right while providing the author with a right to secondary use after one year (§ 38(1) UrhG). This provision has long been overlooked because it is not mandatory and can be overridden by contract. However, examining the history of this provision and comparing it with other provisions strongly suggests that such deviations are strictly limited, if provided for in terms and conditions. German law requires adherence to essential principles of the law when deviations from legal rights are included in general terms and conditions. This arguably covers the balance of interests set out in § 38(1) UrhG, ensuring that exclusive rights are often not unlimited and cannot prevent secondary use by the author, after an embargo period of one year (or, depending on the circumstances, a shorter or longer period).¹⁾

A more specific and seemingly straightforward solution can be found in § 38(4) UrhG, which entitles the author of a publicly funded contribution to a scientific journal to secondary use of their work 12 months after its initial publication. However, this right only covers the secondary use of the accepted manuscript version, and it is unclear whether the page count of the journal may be included. Despite the regulator’s intent to provide greater certainty, the provision still allows for ambiguous interpretation on key issues.

Since both provisions of the (German) UrhG merely restrict an existing right, they do not fall within the scope of exceptions or limitations under the European InfoSoc Directive.²⁾ In other words: The German legislator is entitled, but not obliged, to specify the extent of the right(s) granted to publishers of collections such as journals. This means the legal situation can differ significantly between Member States. However, similar provisions appear to be found for example in

• Austria (§ 37a Austrian UrhG: similar to the German provision),
• France (Article L533-4 Code de la recherche: 6 or 12 months depending on the discipline),
• Netherlands (Artikel 25fa Auteurswet: no specific time frame),
• Spain (Artículo 37(2) Ley 14/2011, as amended by Ley 17/2022: obligation for researchers to open access under conditions), and
• Italy (Art. 4 of Law 112/2013: mixture of first and secondary open access, with an embargo period of 18 or 24 months).

Although many Member States give authors the power to decide on secondary use, some uncertainties and the absence of automatic solutions for the secondary publication after the embargo period may hinder the effective exercise of this right. This could be one explanation for the presumably low secondary use rate (less than 6% in one German study).³⁾

Some funders encourage or even require by their guidelines to make use of their right (e.g., encouraged under DFG 13.2 or required under NIH in the US). In Germany, one federal state and one university have embarked on a field trial, requiring their researchers to exercise their secondary use rights as soon as the embargo ends. This provision was challenged and a decision by the German Federal Constitutional Court has been pending since 2017.

Contract scope with regard to AI as unknown types of use

Publishing contracts and granted usage rights usually cover not only known rights (such as any necessary use for the print and e-book versions of a book), but also related unknown types of uses. Until at latest 2024, one such unknown type of use was arguably the use of the work for the training of generative artificial intelligence (AI) models, or the use as input context for a trained model.⁴⁾ These circumstances give authors significant leverage with regard to past contracts and usage rights.

Firstly, under German law, not all unknown types of use can be covered; they must align with the purpose of the (original) contract. This is debatable for AI use cases, and assuming this is does not apply to contracts for publishing scientific works, it would require the publisher to agree with the author on extending the contract and usage rights. Secondly, even if such AI use cases were covered, § 31a UrhG inter alia requires the publisher to inform the author, who can then revoke such right. More importantly, § 32c UrhG entitles the author to additional remuneration. This could be particularly relevant given that AI training provides for entirely different business cases and significant additional revenue for publishers.

Authors have insofar the opportunity to decide the (AI) future of their old works, which, in most cases, gives them significant leverage compared to new works and contracts covering already any rights including those relating to AI use (see, e.g., the provisions of the German C.H. Beck Verlag)

Right to use the work for AI purposes, in particular Large Language Models

Even if authors granted rights to the publishers, there may in any case remain rights with regard to AI for society in general, including the author, but these are limited to works which are lawfully accessible. However, from the author’s perspective, it may be considered a less powerful or even dangerous tool because it may allow the further use of the work beyond the author’s intent and control. It also comes with even more uncertainty than the right to secondary use.

Such rights derive from the European-wide enacted Copyright exception (Art. 3-4 DSM Directive) allowing text and data mining (TDM). Text and data mining is defined as “any automated analytical technique aimed at analysing text and data in digital form in order to generate information which includes but is not limited to patterns, trends and correlations”. If and to the extent reproductions of works or database extractions are necessary for the purpose of TDM, they can be used for scientific research and – if no machine-readable opt-out has been given by the rightholder – also for other purposes.

Uncertainties have been raised regarding the scope of TDM and the two specific exceptions. Whether the use of a work for the training of generative AI (or the use as context for a prompt) constitutes TDM, despite not being anticipated by the legislator a few years ago, has been debated (see also recently on Verfassungsblog). While there are strong arguments in favour of considering AI training TDM under some circumstances (see Article 53(1)(c) of the AI Act referring to the TDM exception in the context of AI training), it remains to be seen how the Court of Justice of the European Union (CJEU) will rule in the Like Company v. Google case. The details of the research and general TDM exception are also unclear. Arguably, TDM must be linked to a specific scientific research project in order to distinguish clearly between scientific and non-scientific use cases. However, the first German court ruling creates false expectations in this regard by interpreting the research exception as well as the machine-readability requirement of the opt-out too broadly (i.e., natural language is sufficient).⁵⁾

For now, authors can reasonably assume that there are some general exceptions to the use of their work when it comes to AI use (e.g., for training or input context), regardless of the right to secondary use. However, the lawfulness of the use of AI depends on the specific use case and provider (e.g., creating new works vs. copying and imitating old works, providers with zero retention vs. providers using works for own training purposes under their control).

Without prejudice to any future judgements of the CJEU, copyright law provides not only authors but anyone with a range of exceptions regarding the use of AI. However, aside from the use for scientific research purposes, the “rightholder” decides whether to opt out of the TDM use for any purpose. In this context, the original agreement may be relevant.

Summary

“Who owns the work?” – In terms of the scope of the publishing right, secondary use, and AI, the answer is: not only the publishing company, but also the author. In Germany and many other European countries, authors have the right to secondary publication of their works published in journals and other collections (this, however, does not extend to monographs). While specifics of the embargo period and format may be unclear, this right does exist. It is up to the authors to exercise their right and make their work accessible to any institution or person, regardless of disciplines and database subscriptions. Furthermore, when it comes to older works, it is most likely up to the authors to decide whether to allow AI use cases and, if so, under what conditions.

With regard to science, the legislator has already made a wise (?) choice: Anything that is available can be used for TDM and is thus likely to be used in the context of AI – and it is also the authors who decide what becomes available in the first place.