Europe’s AI copyright risk rises without fair use

AI copyright litigation is expanding beyond the United States, where more than 70 AI copyright infringement lawsuits have been active or recently resolved in US and international courts. Major publishers have sued Meta and Google over alleged use of copyrighted books and journal articles to train Llama and Gemini models, while OpenAI, Anthropic, and Microsoft face similar claims. Anthropic settled with authors in 2025 for USD 1.5 billion in Bartz v. Anthropic.

Europe presents a tougher framework for AI developers because Dutch and EU law do not recognize fair use. The DSM Directive’s text and data mining exceptions are limited: Article 3 applies to research organizations, while Article 4 can be blocked through machine-readable rights reservations such as robots.txt. German litigation brought by GEMA against OpenAI has already tested whether memorization and reproduction of protected works exceed those exceptions, and Like Company v. Google (C-250/25) could bring further guidance from the CJEU.

The EU AI Act adds compliance pressure. General-Purpose AI model obligations have applied since 2 August 2025, with fines from 2 August 2026 of up to 3% of global annual turnover or EUR 15 million, whichever is higher. In the Netherlands, the WAMCA and evolving jurisdiction rules could make collective copyright and privacy claims more attractive, especially where claimants use Dutch entities as anchor defendants for broader EU-wide proceedings.