When Germany’s federal and state governments presented their ‘Fortschrittsagenda Kultur + KI’ (Progress Agenda for Culture + AI) in mid-October, they made a major promise: in the age of AI, culture should not merely react, but help shape developments. To this end, collections are to be digitally catalogued and made accessible, and cultural data turned into strategic infrastructure. Yet who decides how these data are used, who pays for their preparation and who benefits from their use remain unresolved questions.
For now, the paper is a shared framework, not a funding programme with fixed budgets. The federal and state governments intend to draw up measures, targets and verifiable outcomes by 2027. The German Cultural Council welcomes the focus on the cultural sector as a whole, but argues that the agenda has yet to adequately address the practical conflicts involved — including copyright, remuneration and the consequences for paid work.
Data are not simply ‘there’
Cultural data can mean very different things: digitised images and sound recordings, catalogue entries, metadata about objects, rights information or digitally documented performances. Such holdings can facilitate research and public engagement. For AI systems, they can also provide material for training models, developing search tools or generating new content. ‘Accessible’ is therefore not a neutral attribute: what matters is to whom data are available, for what purpose and under what conditions.
The policy outline sets the goal of making the core of Germany’s cultural heritage digitally available, accessible through AI and usable across systems. At the same time, it stresses that creative practitioners and cultural institutions should retain sovereign control over their works and data. How this control is to be secured in practice remains unclear. The paper describes neither a standardised approval procedure nor binding licensing models or rules for handling copyrighted works. Nor is there a blanket answer to the question of whether an institution actually holds all the rights to a digitised item: ownership of the object, copyright in the work and rights to a recording are different things.
Who decides — and who pays?
Political responsibility is distributed across the federal system. The federal and state governments jointly agreed the policy outline; local authorities support many museums, archives and libraries. Yet the individual institutions and the bodies responsible for them are crucial to the actual work of digitisation, cataloguing and providing access. The paper outlines joint coordination by a small team drawn from the federal and state governments. Other stakeholders are to be involved in the process; according to the paper, their selection and the form of their participation have yet to be determined.
Funding, too, has yet to be specified. Digitisation means more than scanning an object once. Data need to be described, technically maintained, checked for legal issues and stored for the long term. In its statement of principles, the German Cultural Council points to patchy digitisation, a lack of infrastructure and limited resources in memory institutions. Without sustained funding, smaller institutions in particular could be expected to supply data while having little time or staff capacity to clear rights and maintain those data. This is a foreseeable consequence of the tasks described — not a cost calculation set out in the policy outline.
Participation needs more than a survey
The federal and state governments have announced a wide-ranging survey of cultural associations, institutions, researchers and technology companies. This opens up a route to participation, but does not replace a decision on how different interests will subsequently be weighed. The Cultural Council’s position makes clear what is at stake: it calls for creative practitioners’ perspectives to be included early on and for fair conditions governing the use of their work. In a statement on the European AI strategy, the association describes risks to remuneration and earning opportunities, particularly for self-employed people working alone. These are the positions of an interest group, not a conclusive, evidence-based assessment of the consequences.
For institutions, there is a further question about how benefits are shared: if publicly funded collections provide data and private providers derive commercial benefit from them, what constitutes an appropriate return? The agenda commits to fair remuneration and a share in the added value, but does not yet establish a procedure for this. It is equally unclear how non-commercial research, public engagement and commercial AI training are to be distinguished from one another.
A framework, not yet a guarantee
The Progress Agenda can encourage shared standards and collaboration. But it cannot, on its own, resolve the copyright questions that the paper explicitly assigns to other political processes. Nor does it yet specify funding amounts or a binding division of costs between federal, state and local governments.
The next step will therefore determine the nature of the undertaking. Will smaller institutions and creative practitioners have a say in the rules? Will there be money for ongoing data maintenance, rather than just technical pilot projects? And will it remain clear who is using data and for what purpose? Only when these questions have been answered will the promise of cultural data sovereignty become a verifiable practice. Until then, the paper is primarily a brief for further work — with important goals, but as yet no firm commitments to those expected to catalogue data, make them accessible and safeguard the associated rights.
Original reviewed by the editors · 26 October 2026
