Policy resources
Analysis and reference material on text and data mining, the CDSM Directive and AI development in Europe.
We are building an overview of how Articles 3 and 4 of the CDSM Directive have been transposed across the Member States, together with relevant national case law and guidance. The aim is to make divergences easier to identify and to support a more consistent application of EU law.
Working notes on the design of interoperable, machine-readable rights-reservation mechanisms under Article 4(3) CDSM: what a signal should express, how it can be discovered and interpreted, and how conflicting or invalid signals should be handled.
Research reports, policy briefings and Coalition documents on copyright and AI.
The Lisbon Council · Economic analysis ·
Bartłomiej Biga compares TDM rules across the EU, US, Japan and Singapore, assessing their effects on innovation, transaction costs and economic growth. The report argues for retaining the EU’s opt-out framework with clear, machine-readable rights reservations. Financially supported by Meta.
Oxford Economics · Research report ·
A comparison of copyright frameworks and AI ecosystems across seven Asia-Pacific countries. The report finds that clearer, more flexible AI-training rules are associated with stronger research and innovation, while stressing the importance of infrastructure, talent and investment. Commissioned by Google.
Brian Williamson · Research report ·
Brian Williamson examines how AI can support European productivity, science, industry and culture. He argues that a stable TDM exception and access to existing knowledge are essential to Europe’s AI capabilities, while distinguishing training inputs from copyright protection of outputs. Independent report funded by Google.
ECIPE · Policy Brief No. 09/2026 ·
Oscar Guinea, Dyuti Pandya and Vanika Sharma assess the wider economic trade-offs of restricting AI training data in Europe. The brief recommends preserving the TDM exception, keeping rights reservations machine-readable and avoiding mandatory licensing that could disproportionately burden start-ups.
Manifesto (PDF)
