Artificial Intelligence and Copyright: A Danish Perspective

Terese Foged
Bech-Bruun, Copenhagen
tef@bechbruun.com
    

Introduction

The tension between generative artificial intelligence (AI) and copyright has moved rapidly from theory to policy and litigation. Denmark offers an instructive vantage point, combining a distinctive licensing tradition with active engagement in the European debate on the text and data mining (TDM) exception.

This article outlines three current Danish developments: the recommendations issued by a government-appointed expert group, the emerging use of the extended collective licence for AI and the first AI copyright cases now before the Danish courts.

Recommendations from the expert group on copyright and AI

In September 2025, a government-appointed expert group on copyright and AI published its report, setting out a series of policy recommendations aimed at strengthening the position of rightsholders in the age of generative AI.[1]  Three recommendations stand out.

First, the group recommends that the Danish government work at EU level to secure effective transparency in relation to the training data used by AI systems. Transparency is treated as a precondition for any meaningful enforcement of rights: rightsholders cannot assert or license their works if they cannot establish whether, and to what extent, those works have been used. Some members went further, advocating a reversed burden of proof and a presumption that copyright-protected content has been used as training data, so that the onus would fall on the AI developer to show that protected works have not been used.

Second, and more fundamentally, the group recommends amending the TDM rules to improve rightsholders’ ability to prevent their works from being used in text and data mining. The current EU framework, transposed through Articles 3 and 4 of the Directive on Copyright in the Digital Single Market (the ‘DSM Directive’), permits commercial TDM unless the rightsholder has expressly reserved use (a so-called ‘opt out’). The group considers this default unsatisfactory and recommends that, in connection with the European Commission's planned evaluation of the DSM Directive in 2026, Denmark puts forward a proposal requiring the rightsholder’s express consent for TDM. This would move the situation from the present ‘opt-out’ model to an ‘opt-in’ model, shifting the burden from rightsholders, who must currently take active steps to protect their works, to AI developers, who would have to obtain affirmative permission before carrying out training activities based on protected material.

Third, the group addresses enforcement, considering the existing tools for resolving rights disputes inadequate. It recommends examining a compulsory ‘final offer arbitration’ model, initially as a pilot scheme for press publications. Under that model, the parties must engage actively in negotiation and mediation. If those avenues fail, each submits a final offer to an independent arbitral body, which selects the more reasonable proposal. The design is intended to encourage balanced offers and deliver swift, binding outcomes, particularly where, as with press publications, the content serves an important democratic and informational function. If pursued legislatively, the group considers the rules would be most effective if adopted at EU level.

The role of the extended collective licence

A second strand of the Danish discussion concerns the extended collective licence. Originally a Nordic model, it allows an agreement between a user and a representative collective management organisation to be extended by law to cover the works of non-member rightsholders in the same category, subject to safeguards, including an opt-out right.

Legislative remarks and related preparatory materials indicate that the extended collective licence can be used to clear rights for AI purposes. It is seen as particularly relevant where a use cannot proceed under the TDM rules, including where rightsholders have opted out of TDM, so that the statutory exception is no longer available for their works.

This connects the opt-out and opt-in debate to practical licensing. Opt out from TDM appears increasingly widespread: many rights organisations encourage their members to opt out, and large undertakings, including major media companies, have done so. As opt out becomes the norm, the TDM exception covers an ever-smaller share of the relevant repertoire, and a positive licensing route becomes essential if lawful AI uses are to continue. The extended collective licence offers precisely such a route, clearing large bodies of works efficiently where individual clearance would be impractical and the TDM exception does not apply.

This is no longer theoretical. In December 2025, Denmark saw its first extended collective licence approvals in the AI field: supplementary agreements on the use of AI, supplementing the primary and lower secondary school agreement and the workplace licence.[2]  This demonstrates that the extended collective licence can be adapted to authorise AI-related uses within defined sectors, and they may serve as a template for further AI licensing.

The first AI copyright litigation in Denmark

The third development is litigation. Danish rightsholders and rights organisations are now testing the boundaries of copyright protection against AI service providers before the Danish courts.

First, Koda, the collective management organisation representing composers, songwriters and music publishers, has sued the United States AI music service Suno before the Danish courts, raising the question of whether and how the training and output of a generative music service engage Danish and EU copyright rules.[3]

Second, the Danish media organisation DCPMO has sued OpenAI and, separately, LinkedIn,[4] in both cases before the Danish courts.[5]  These coordinated actions assert press publishers’ rights against the use of their content in connection with AI services.

The Danish government has also intervened in the principal EU case on this question, Like Company v Google Ireland, Case C-250/25, concerning Google’s use of media content by chatbots.[6]

Taken together, these cases signal that Danish rightsholders are not content to await legislative reform. They seek, through litigation, to clarify the application of existing copyright rules to generative AI, including the scope of the TDM exception and the effect of opt-out reservations.

The wider Danish policy context

These developments sit within a broader policy push. During the Danish EU Presidency, at an informal ministerial meeting in Copenhagen on 4 November 2025, the Danish Minister for Culture convened European ministers for culture and media, who issued a joint declaration on the necessity of culture and media as a safeguard for our European democracies.[7] The declaration warned that the use of AI in producing and distributing cultural and media content raises ethical, societal and economic questions and questions of reliability, and stressed that culture, cultural heritage and media policies must form a vital part of European collaboration to protect democracy.

At the domestic level, rightsholder organisations have likewise pressed the issue, including through an open letter urging that copyright and AI concerns be taken into account during government negotiations.[8]

The Danish influence extends to the Council of Europe. In April 2026, the Parliamentary Assembly’s Committee on Culture, Science, Education and Media adopted a report on copyright enforcement in the AI environment, prepared by a Danish rapporteur.[9] Its draft resolution calls on Member States, among other measures, to clarify that the TDM exceptions do not apply to AI training, to require AI providers to disclose the training data, to introduce a presumption of training on copyright-protected material where transparency requirements are not met and to support fair remuneration and collective licensing. It also points to a Danish final offer arbitration model as a potential enforcement tool, echoing the pilot scheme proposed by the expert group.

Together, these initiatives show Denmark seeking to shape the European response to AI through cultural and democratic policy at both the EU and wider European level.

Conclusion

Denmark illustrates three complementary responses to the challenge that generative AI poses for copyright.

At the policy level, the copyright and AI expert group has recommended greater transparency in regard to the use of training data, with some members advocating a reversed burden of proof; a shift from an opt-out to an opt-in model for TDM, to be advanced in connection with the Commission’s 2026 evaluation of the DSM Directive and a pilot final offer arbitration scheme for press publications.

At the licensing level, the extended collective licence is emerging as a practical clearance mechanism, as the first approvals in December 2025 show.

At the enforcement level, rightsholders are pursuing pioneering litigation against AI service providers, and the government has intervened in the key EU case concerning use of media content in chatbots, Like Company v Google Ireland.

Danish influence is also visible at the Council of Europe, where a Danish-led report has called for far-reaching reform, including regarding the arbitration model.

Together, these developments make Denmark a jurisdiction worth watching as the European response to AI and copyright continues to take shape.