Blog post by Maja Bogataj Jančič and Laura Pipan, Open Data and Intellectual Property Institute ODIPI
On 10 November 2025, members of European public AI research laboratories and AI model builders addressed an open letter to the European Commission, proposing measures to increase legal certainty for public AI model builders in Europe.
In this open letter, these stakeholders expressed concerns about the lack of a clear framework for AI development in Europe. Urgent action is needed to ensure Europe remains a competitive and legally secure AI hub. Global AI progress is accelerating through major investments in the US, China, and other regions, and Europe’s ability to benefit depends on clear and favourable rules for training and publishing open-source AI models. Current copyright uncertainty is slowing progress and risks Europe investing heavily while others lay the foundations of AI. They urged the Commission to confirm that current EU law permits AI model training on lawfully accessible datasets to foster innovation.
Key Action Points to Deliver the Next Generation of AI “Made in Europe”
They highlighted four critical points:
I. Legal Clarity for a Competitive European AI Model
Europe already has a legal basis for AI training under Articles 3 and 4 of the Copyright in the Digital Single Market (CDSM) Directive 2019/790, covering text and data mining (TDM). However, uneven implementation and restrictive interpretations across Member States threaten model training. Competitive AI requires broad, diverse datasets, including web-crawled content; without legal clarity, public research initiatives, startups, and SMEs will fall behind global competitors.
II. Establishing a Safe Harbour for AI Model Training
The signatories urged the Commission to clarify that TDM exceptions already fully allow AI model training on lawful data, respecting rightholders’ opt-outs where appropriate. In addition, sharing derived datasets prepared for AI training and collaborative TDM infrastructures should also be legally recognised, harmonising rules across Member States and facilitating cross-border research. A safe harbor would provide legal certainty for European AI development.
III. Aligning Investments with European AI Strategies
The European Union has already promised significant investments in AI through its EU AI Continent Action Plan and Apply AI Strategy, both of which aim for a competitive, sovereign, and trustworthy AI ecosystem. A safe harbour for public AI developers would be critical for realising the potential of this spending – legal clarity is as critical as computing power or research funding, enabling strategic investments in AI Factories and AI Gigafactories.
IV. Europe’s Opportunity and Call to Action
European public AI research laboratories and AI model builders are ready to collaborate on open, competitive AI models reflecting European values. Clarifying what counts as legal AI training signals that the EU not only regulates but enables AI, turning strategic investments into frontier-level models, strengthening open-source ecosystems, and attracting talent.
The signatories stand ready to deliver the next generation of AI “Made in Europe.” Establishing a Safe Harbour now is essential to secure Europe’s role as a builder and shaper of the AI age.
Let’s Invest in Public AI
Europe should invest further in the development of Public AI precisely because of the possibility of cooperation between research and cultural institutions, which is already enabled by Article 3 of the CDSM Directive. The Directive provides a legal basis for TDM and thus the legal training of AI models on all lawfully accessible works without the need for special permissions from the rightsholders, when it comes to scientific research purposes or non-commercial research purposes. The entire process of developing Public AI, from data acquisition to training and publishing open-source models, is legal according to Article 3 of the CDSM Directive, which is also confirmed by legal opinions, including from Prof. Dr. Malte Stieper, who recognises the right of libraries and research organisations to train models from their collections and to participate in joint projects.
Open source LLM builders rarely speak with one voice, but they are now asking the European Commission to provide legal clarity, as many projects face uncertainty and pressure from rightsholders, which hinders European progress and threatens the strategic objectives outlined in the Apply AI Strategy or AI Continent Action Plan. They are explicitly calling for an interpretative communication confirming that Article 3 of the CDSM Directive applies to the training of public AI models. Europe should not only regulate AI but also enable it. As Paul Keller from Open Futures has argued, a balance should be maintained between research and rights holders, with the possibility of compensation for commercial use of AI.
KR21 National Coordinator for Slovenia Dr. Maja Bogataj Jančič participated in the various discussions and events (here, here, here, etc.) focusing on the copyright legal basis for building open source national LLMs and has provided support to the LLM builders in voicing their concerns to the Commission.
24 November 2025
