On 1 June, COMMUNIA brought together policymakers, researchers, Public AI developers and cultural heritage professionals in Brussels for the event Copyright Liability and the Knowledge Ecosystem. The event explored a question that is becoming increasingly urgent for Europe’s knowledge institutions: how can public-interest actors exercise the rights granted to them by copyright exceptions when legal uncertainty makes relying on them increasingly difficult?
The event opened with an update on the European Commission’s ongoing work on copyright from Lars Frietman (DG CONNECT, European Commission), including the DSM Directive survey and the targeted initiative for a better copyright for European creativity and innovation. It also marked the launch of a new IViR research publication examining legal uncertainty and risk mitigation in the European copyright framework. The day concluded with reflections from MEP Sergey Lagodinsky on the broader societal implications of AI, the need for a measured approach to future regulation, and the importance of supporting public interest institutions and Europe’s digital sovereignty objectives.
You can now watch the full recording of the event below.
Public AI developers and copyright: where legal uncertainty starts
The first panel examined how copyright uncertainty affects the development of public interest and open source AI systems in Europe.
The discussion was introduced by Paul Keller (Open Future/COMMUNIA), who presented an ongoing research project on sources of legal uncertainty affecting public-interest AI development. The panel discussion brought together Jan Hajič (Charles University Prague) and Anna Leschanowsky (Fraunhofer IIS / Soofi – Sovereign Open Source Foundation Model Initiative), and was moderated by Paul Keller.
The discussion highlighted a growing mismatch between the European Union’s ambition to foster sovereign and public-interest AI ecosystems and the practical realities of the current copyright framework. While the DSM Directive introduced text and data mining (TDM) exceptions intended to support innovation, participants stressed that these provisions often fail to provide sufficient legal certainty for publicly funded and open source AI initiatives.
A central theme was the relationship between the DSM Directive’s research TDM exception (Article 3) and the broader TDM exception (Article 4). Public interest AI projects increasingly develop models that are intended to be openly shared and reused by European SMEs, researchers, and public administrations. Yet uncertainty about whether downstream commercial reuse may affect eligibility under Article 3 pushes many open source projects toward the more restrictive Article 4 framework.
Panelists also discussed the substantial operational costs generated by legal uncertainty. Public-interest developers often invest significant resources in copyright compliance, metadata verification, opt-out detection, and legal review. These burdens are particularly acute for smaller actors and projects working with less-resourced European languages. In some cases, uncertainty leads developers to adopt precautionary measures that go beyond what the law requires, limiting data sharing and reproducibility practices that are essential to open science.
Cultural heritage institutions and copyright: where legal risk should end
The second panel shifted the focus to cultural heritage institutions.
The discussion opened with a presentation by Martin Senftleben (Institute for Information Law (IViR), University of Amsterdam), who presented new research on legal risk mitigation measures for knowledge institutions. He was joined by Dyveke Sijm (Royal Danish Library) and Stephen Wyber (IFLA) for a panel discussion moderated by Teresa Nobre.
Participants noted that one of the main obstacles faced by these institutions is no longer the absence of copyright exceptions. Instead, the challenge lies in the difficulty of relying on existing exceptions with sufficient legal certainty and operational confidence.
The panel also explored how uncertainty creates a powerful chilling effect. Libraries and cultural heritage institutions frequently adopt highly risk-averse approaches, not because copyright law clearly prohibits a particular activity, but because the potential legal consequences of relying on copyright flexibilities remain difficult to assess. This often results in defensive compliance, where institutions refrain from lawful activities or accept restrictive contractual terms rather than challenge them.
Against this backdrop, participants argued that future copyright reform should focus not only on creating exceptions, but also on ensuring that institutions can rely on them in practice. The discussion highlighted a range of legal risk mitigation measures proposed in the IViR study, including safeguards against unfair contractual terms imposed on public interest institutions, clearer rules on applicable law and jurisdiction, and limitation-of-liability protections for institutions acting diligently and in good faith.
From rights on paper to rights in practice
Although the two panels focused on different sectors, they converged on a common message: legal uncertainty imposes real costs on public-interest actors across the knowledge ecosystem.
Whether developing open-source AI models, preserving cultural heritage collections, or enabling access to knowledge, institutions increasingly face situations where the law appears to permit an activity, yet the associated legal risks make reliance on that permission difficult in practice. Addressing this challenge will require more than expanding copyright exceptions. It will also require mechanisms that reduce legal risk, increase predictability, and allow public-interest actors to operate with confidence.
