At XX CODAIP on 4 November 2026, Professor Thomas Hoeren explored the boundaries of AI training under European copyright law. His lecture examined text and data mining, the LAION litigation, and the GEMA proceedings against OpenAI and Suno, alongside Advocate General Szpunar’s Opinion of 3 November in Like Company (C-250/25).
Szpunar considers LLM training generally capable of benefiting from the text and data mining exception, although he proposes declaring the training questions inadmissible in this case. His principal focus is real-time retrieval and reuse: chatbot answers may infringe press publishers’ rights when they reproduce recognisable expression to an extent that substitutes for reading the original publication. Mere facts remain unprotected. The Opinion is not binding; the Court’s judgment is still pending.
A central theme of the lecture was the distinction between collecting training data, memorising protected expression and generating outputs. Each stage raises different legal questions—and understanding the technology is essential to answering them.
#Copyright #ArtificialIntelligence #CODAIP
