We need an open, transparent, and creator-centric EU AI strategy for the cultural sector.
Open Nederland is the national network of creators, advisors, legal experts, researchers, and entrepreneurs who believe that knowledge, culture, and technology should be freely available to study, modify, build upon, and share. The association is committed to a digital society where information can be made openly available to serve as a building block for innovation and creativity. Open Nederland advocates for open data, open source software, digital sovereignty, and a robust public domain. Additionally, Open Nederland serves as the official Dutch Chapter of Creative Commons, promoting the use of open licenses and open standards across education, science, culture, enterprise, and government.
-
Public Money, Public Ownership Whenever EU public funding supports the development or implementation of AI systems, it must mandate that all the results are made fully available to the public. This requires AI models to be released as open models (including documentation, training data, provenance, training methodologies, and contributor information) and any underlying software as open source. Adopting the “Public Money, Public Code” principle - rooted in Free and Open Source Software - prevents vendor lock-in, ensures transparency, and allows creators, SMEs, and public institutions to freely study, reuse, and build upon publicly funded innovations.
-
AI That Works for Creatives: Democratising Access AI development is currently dominated by large technology conglomerates, leaving cultural SMEs, non-profits, and independent creators dependent on proprietary solutions. AI must be democratised. EU policy should ensure that AI technologies are accessible, adaptable, and affordable for cultural SMEs and independent creators. Policy must prevent digital monopolies and foster a genuinely diverse creative ecosystem capable of contributing to and expanding culture, the economy, and society at large. Additionally, AI infrastructure that works for governments can also work for creatives, seek synergies with EU-funded or cross national infrastructrural initiatieves, like EU-EDIC.
-
Build capacity through public sector institutions EU policy should build capacity for AI literacy, skill development, and equitable access to AI technologies through existing collaborative public networks, such as cultural heritage institutions and libraries. By deploying open-source tools, training programs, and cross-border datasets, the EU can ensure that public institutions lead an ethical AI transition aligned with European values.
-
Protect the Public Domain and Prevent New Sui Generis Exclusive Rights What belongs to the public domain must remain in the public domain. The EU must prevent the creation of new exclusive or neighboring rights on digital reproductions of public domain works held by heritage repositories. Furthermore, purely AI-generated outputs without human creative input must remain uncopyrighted, preserving the public domain as a vital foundation for future human creativity.
-
Respect and Enforce Open Content Requirements Open Content repositories (such as Wikipedia, Wikidata, and Wikimedia Commons) form a substantial portion of the datasets used to train generative AI models. AI developers must strictly comply with open licenses (e.g., Creative Commons). Attribution requirements must be strictly enforced, source metadata (provenance) preserved, and open-licensing terms respected across all training datasets and at inference time.
-
Safeguard Text and Data Mining (TDM) Exceptions for Research and Culture Measures introduced to protect rightsholders must be balanced with the rights of artists, creators, individuals, researchers, and cultural institutions. Legal or technical restrictions must not undermine or restrict existing Text and Data Mining (TDM) exceptions, which remain essential for scientific research, data analysis, creative endeavours, and cultural innovation.
-
Direct Remuneration to Creators, Not Corporate Rightsholders Rather than relying solely on traditional copyright mechanisms — which often result in corporate value capture, market throttling, and lower-quality AI models — the EU should explore remuneration models that directly benefit human creators rather than corporate rightsholders. While rightsholders manage commercial licensing, generative AI fundamentally impacts the creative labor of artists and creators. Especially within open-content ecosystems, where value is generated through the act of creation rather than distribution, compensation mechanisms must directly reward human effort and creative practice.
-
Privacy Instead of IP Privatisation Broadening intellectual property rights is the wrong approach to address the risks posed by deepfakes and unauthorized digital replicas. Personal attributes — such as a person’s likeness, voice, or identity — are matters of fundamental privacy rights, not property ownership. Legal interventions must strengthen privacy and data protection frameworks rather than privatising human identity through expanded IP monopolies.
-
Radical Transparency in AI Training Data Meaningful oversight of AI systems requires total transparency regarding training datasets. The EU should mandate standardized, open technical mechanisms—such as publishing cryptographic hashes of media files used in training sets (similar to mechanisms considered for centralized TDM opt-out databases). This allows creators, rightsholders, and researchers to independently verify whether specific works were used to train a model, establishing a trustworthy technical foundation for compliance and accountability.
Explicitly supported by
- Open Nederland
- Creative Commons Nederland
- Open state Foundation
Levend document
Dit is een levend document. Leden van Open Nederland kunnen dit document aanscherpen en uitbreiden.
Bewerk dit standpunt op GitHub