As part of the Digital Omnibus Proposal, the Commission proposes to integrate the provisions of the Open Data Directive (2019/1024) into an expanded Data Act. While the Commission proposal argues that the changes introduced through this process are merely technical in nature, this process provides the opportunity to review and fix parts of the existing acquis that are not working in practice.
One such provision is Article 12(3) of the Open Data Directive, which governs exclusive arrangements arising from public-private partnerships in the context of the digitisation of cultural heritage. As we explain in this blogpost, its core protection — that exclusivity periods “in general” shall not exceed ten years — has failed in practice.
The problem: 2 million PD books subject to exclusive restrictions
Article 12(3) had been introduced in part to rein in digitisation agreements between cultural heritage institutions and commercial entities that provide the digitisation partners with exclusive rights with regard to the bulk use of the digitised works for commercial purposes. Prior to the introduction of this provision, digitisation agreements had been concluded that either granted such rights in perpetuity or limited them for periods of up to 15 years. The legislative intent behind Article 12(3) was to regulate such practices and limit exclusive arrangements to no more than 10 years (including retroactively).
Despite this very targeted intervention by the EU legislator, today there are no known digitisation agreements between a major commercial digitisation partner and a European cultural heritage institution where exclusive provisions are capped at ten years. Despite the fact that a large number of agreements concluded between European national libraries and Google were concluded in the early 2010s, there is no evidence — or reason to believe — that these agreements have been reviewed after a period of 10 years. In addition, almost all of these agreements remain private, and library partners are prevented by NDAs from discussing their provisions in public. Even worse, some of the early library partners are still bound by agreements that grant exclusive access in perpetuity.
The consequence is concrete: a large number of digitised public-domain books from European library collections cannot be made available to any party other than the original commercial digitisation partner for the purposes of AI training and other computational uses. Based on what we know about the starting date of existing digitization agreements, we estimate that at least 2 million books are currently subject to exclusivity restrictions that would not apply if the ten-year norm in Article 12(3) were treated as a binding ceiling. This situation significantly limits libraries from contributing to the objective of the 2025 Data Union Strategy of making 30 million+ cultural heritage works available for AI development. It also places European AI developers at a direct structural disadvantage relative to incumbents who benefit from these arrangements.
The fix: amending the Digital Omnibus proposal
The Digital Omnibus proposal, which would transpose Article 12(3) of the Open Data Directive into Article 32k(3) of the amended Data Act, provides a timely opportunity to correct this weakness and provide more clarity.
Article 12(3) reads as follows:
By way of derogation of paragraph 1, where an exclusive right relates to the digitisation of cultural resources, the period of exclusivity shall in general not exceed 10 years. Where that period exceeds 10 years, its duration shall be in accordance with applicable Union and national law subject to review during the 11th year and, if applicable, every seven years thereafter.
We propose to turn the general rule that “the period of exclusivity shall in general not exceed 10 years” into an effective maximum ceiling for exclusive arrangements. We further propose to remove the review clause that has never been used and provides no meaningful check on long-running or uncapped agreements. In addition it could be considered to make it mandatory to make the arrangements public on the website of the cultural heritage institution that has entered into such an agreement.
Why This Matters Now
While this situation has persisted for some years, it has become more salient given the increasing demand for digitised cultural heritage from AI model developers.
The proposed amendment is aligned with EU AI policy. The European Data Union Strategy and the Commission’s own objectives for making cultural heritage content available as AI training data are directly undermined by the current failure of enforcement. Strengthening Article 12(3) is a straightforward legislative step toward meeting those objectives. The proposed amendment fits within the regulatory objective of simplifying the digital rulebook and provides an opportunity to unlock a meaningful amount of valuable training data for EU AI developers on a short timeline.
The proposed amendment would provide legal clarity for libraries. Libraries holding digitised collections have been reluctant to assert their rights under Article 12 precisely because the current text offers only a general norm, not a binding obligation. A hard ceiling would give institutions a clear legal basis to act. This is especially important given that the demand for digitised cultural heritage works as AI training data has led to a new wave of digitisation partnerships between European libraries and non-EU incumbents that risk further entrenching their competitive advantage in AI development.
The proposed amendment would not create new obligations on private partners. The amendment does not impose retroactive liability or materially change the terms under which affected parties have operated since the adoption of the Open Data Directive. It does, however, clarify that new and existing agreements must come into compliance — and that ongoing non-compliance is not consistent with EU law.
