Participants:
Paul Keller – Policy Director at Open Future, co-founder of Communia Association
Oana Nasui – Cultural researcher and project coordinator
Origins of Communia and the Public Domain Movement
Paul Keller explains that Communia Association emerged from an EU-funded research project that brought together activists from the Creative Commons communities across Europe (Netherlands, Germany, Greece, Poland, Spain, Catalonia) around 2005-2006. At that time, almost 20 years ago, many people felt the internet had enormous potential for enabling access to information, but much of that potential remained unrealized. This was before music subscription services, streaming platforms, or widespread digital access – a period when young people wanted online access to music and information, but faced significant barriers.
The three-year EU-funded project built a strong network of people who cared about openness and access to information. After the project ended, they used part of the funding to establish an international association under Belgian law – the Communia Association – which still exists today with the mission to advocate for the interests of the public domain and access to information.
Defining the Public Domain – A Broad Conception
Keller articulates Communia’s comprehensive understanding of the public domain: information that isn’t owned by anyone and is available for everyone. This includes old information no longer protected by copyright due to term expiration, pure information that isn’t creative expression and therefore not protected by copyright, and voluntarily shared content under Creative Commons licenses (like Wikipedia).
This broad conception emphasizes that free access to as much information as possible is fundamentally important for society. Communia operates primarily in Brussels (hence the Belgian legal structure) and at WIPO (World Intellectual Property Organization) in Geneva, advocating for these principles at both EU and international levels.
The Public Domain Manifesto and Policy Framework
Communia’s guiding principles are articulated in the Public Domain Manifesto (published around 2009) and accompanying policy recommendations, which were updated two years ago. Over 12-13 years of existence, Communia has become a frequent and respected participant in copyright policy discussions in Brussels, providing a crucial counterbalance to the numerous organizations representing specific creator groups, creative industries, and rights holders.
Keller emphasizes that most discussions about access to information are filtered through the lens of copyright, which essentially defines the boundary between what is protected and what belongs to the public domain. Historically, very few voices – mostly academics – understood the importance and value of the public domain in these policy debates.
Copyright as a Territorial vs. Global Challenge
A central theme emerges: copyright is fundamentally territorial, designed for physical media like books and phonographic records that could be controlled within specific jurisdictions. The law was developed in response to technologies that were physical in nature and could be “pinned down” geographically.
However, modern digital technologies and AI operate globally on the internet, creating a profound mismatch. As Keller states: “Everything is global except the regulatory framework.” This territorial nature of copyright creates significant enforcement challenges when content, platforms, and AI models circulate freely across borders.
The EU Copyright Directive and Cultural Heritage Digitization
Nasui introduces her research on the EU Copyright Directive, particularly focusing on how it impacts cultural heritage institutions.
This provision is crucial for cultural heritage institutions seeking to digitize their collections. The directive creates a mechanism where cultural institutions can apply for permissions to digitize these orphan works or out-of-commerce materials, which would otherwise remain inaccessible due to copyright restrictions. This represents an attempt to balance copyright protection with the public interest in preserving and accessing cultural heritage.
The Orphan Works Problem
The discussion delves into the challenge of orphan works – copyrighted materials where the rights holder cannot be identified or located. This is a massive problem for digitization projects, as institutions cannot obtain permission to digitize and make available works whose owners are unknown.
The EU directive attempts to address this through “diligent search” requirements, allowing institutions to use orphan works if they can demonstrate they’ve made reasonable efforts to locate rights holders. However, Keller notes this remains a complex and resource-intensive process that continues to impede digitization efforts.
The 25-Year Rule and Commercial Availability
An important technical detail emerges: the directive includes a 25-year rule for determining when works are considered “out of commerce.” If a work has not been commercially exploited for 25 years, it may qualify for the special digitization provisions.
This rule attempts to balance the interests of rights holders (who may want to commercially exploit their works) with the public interest in accessing cultural heritage that has effectively disappeared from the market. However, implementation varies significantly across EU member states, creating inconsistencies in how cultural institutions can operate.
The Power Imbalance: Creators vs. Tech Giants
Keller identifies a fundamental power imbalance in copyright discussions. While creators and rights holders organize at the national level (writers’ unions, visual artists’ associations in individual countries), they face global corporations like Google, Microsoft, and OpenAI operating across borders with vastly greater resources.
Even when all organized writers or visual artists in larger EU countries band together, they remain “comparatively small as an organization” compared to tech giants. This structural imbalance makes traditional copyright enforcement increasingly challenging and raises questions about whether copyright can effectively regulate global digital platforms.
AI and Copyright – The Training Data Debate
The conversation turns to one of the most contentious contemporary issues: whether using copyrighted material to train AI models constitutes copyright infringement. Keller outlines the fundamental debate:
Position 1 (Many creators and rights holders): Using copyrighted works to train AI without permission is copyright infringement. They argue that AI companies are building profitable businesses on the unauthorized use of creative works, and creators should be compensated.
Position 2 (Many in the tech/open access community): Training AI models is analogous to how humans learn – by reading, viewing, and absorbing existing works. From this perspective, it should be considered fair use or covered by existing exceptions for text and data mining.
Keller notes there’s a spectrum of positions rather than two absolute camps, with many nuanced views in between. The debate touches on fundamental questions about the nature of learning, creativity, and what constitutes “copying” in the digital age.
Text and Data Mining (TDM) Exceptions
The EU Copyright Directive includes provisions for text and data mining – allowing researchers and others to analyze copyrighted materials computationally. However, rights holders can opt out of these provisions, creating uncertainty about what materials can be used for AI training.
This opt-out mechanism reflects the ongoing tension between enabling technological innovation and research while respecting rights holders’ interests. Different EU member states have implemented these provisions differently, adding to the complexity and creating a patchwork of regulations across Europe.
The Getty Images Case – Jurisdictional Limits
Keller discusses the Getty Images vs. Stability AI lawsuit as an illustrative example of copyright’s territorial limitations. Getty Images sued Stability AI, presenting evidence that the AI system produced images containing Getty’s watermark – suggesting the training dataset included Getty images.
However, Getty had to retract part of their claim when it became clear the AI training occurred outside the United Kingdom, where the lawsuit was filed. UK copyright law only applies within UK territory, making it irrelevant what happened elsewhere during the training process.
This case demonstrates that instinctively, most people would agree that taking images without permission and using them to build a profitable system seems wrong. Yet the territorial nature of copyright law makes it extremely difficult to enforce such intuitions in a global digital environment.
The Mismatch Between Global Technology and Territorial Law
Keller emphasizes that AI models circulate globally on the internet, which is also global, but the regulatory framework remains stubbornly territorial. Copyright has “real limits as a system that is super territorial” when dealing with technologies that know no borders.
The underlying mechanisms powering copyright were developed in response to physical technologies – printing books, creating phonographic records – that could be controlled within specific geographic boundaries. These mechanisms are increasingly inadequate for regulating digital information flows and AI development.
From Individual Creativity to Big Data Business
Nasui draws an important distinction: before AI, individuals with internet access could become “content producers,” creating collages and remixes from bits of existing material. This grassroots creativity was generally non-commercial and fell into legal gray areas.
However, AI operates at a completely different scale – “billions of big data” – and is explicitly “mission-based” with clear commercial objectives. Companies are building huge businesses by aggregating vast amounts of data, fundamentally changing the stakes of the copyright debate. The scale and commercial nature of AI development make it qualitatively different from individual creative experimentation.
The EU as the Most Viable Regulatory Level
Despite critiques of EU bureaucracy, Keller argues that the European Union represents the most viable level for effective regulation. While some may view the EU as overly bureaucratic, it operates on “a set of shared values” with “common agency” that makes collective action possible.
Global regulation would be ideal, but “we don’t have the structures” for effective global governance in this area. Therefore, Europeans must think strategically about “how can we organize ourselves in a way that this imbalance of power is not as big as it is currently.” The EU, despite its imperfections, offers the best available framework for balancing the interests of creators, the public, and tech companies.
Short-term Solutions in Exponential Times
Nasui concludes by emphasizing the need for short-term, practical solutions given AI’s exponential growth trajectory. We “can hardly measure and see where it’s going,” making long-term legislative solutions potentially obsolete before implementation.
Public discussions in virtual spaces may not result in immediate legislative resolutions, but they can “open up our brains” to find workable approaches. Understanding the system and network of copyright, digitization, and AI governance helps stakeholders develop better proposals for navigating these challenges.
The conversation emphasizes that thinking at higher organizational levels (EU rather than national) increases the “chance of success” in creating effective frameworks that can address the power imbalances inherent in the current system.
Justice and Systemic Balance
The dialogue closes with Nasui’s concern about “justice” – ensuring fair treatment for creators, cultural operators, and the public interest in accessing information. Keller’s response emphasizes building “strong enough” organizational structures to reduce the current power imbalance.
This isn’t about creating “countervailing” opposition, but rather finding ways to organize that allow for more balanced negotiations between vastly unequal parties – individual creators and national organizations on one side, global tech giants on the other.
Final Note: This dialogue provides crucial insight into the complex intersection of copyright law, cultural heritage digitization, AI development, and the structural challenges of governing global technologies through territorial legal frameworks. It highlights the urgent need for coordinated action at the EU level to address power imbalances and protect both creator rights and public access to information in the AI age.
Context: This professional dialogue is part of the “Beyond Creation” project, co-financed by the Administration of the National Cultural Fund (AFCN) and produced by the Formare Culturală platform.
