GEMA filed a lawsuit against Suno in early 2025, focusing on the "memorization" and highly similar outputs of its model concerning specific German and international pop songs. The Munich court applied the logic from a previous case involving OpenAI lyrics, clearly stating that both training and output constitute actionable reproduction, breaking through some of the applicability of the U.S. fair use defense in Europe.

In terms of capital and compliance pathways, Suno recently completed significant financing but faces mandatory licensing and potential high damages in the European market. Rights holders are using this ruling to push for industry-wide negotiations, motivated by the desire to convert "free training data" into chargeable licensing income while raising the entry barriers for newcomers.

Comparing the music industry's early sampling and digital download protection paths, as well as a series of lawsuits in the text domain against large language models, the Suno case marks a turning point from "technical exemption to mandatory licensing" for generative AI.

Essentially, this represents a shift in regulation and pricing power: when the court confirms that memorization in model weights constitutes infringement, copyright extends from "output control" to "training source control," forcing AI companies to reprice data acquisition costs and reshape the profit distribution of the entire generative music industry chain.

ABAB News · Cognitive Laws

The free lunch of training data will eventually end
Memorization is reproduction, and reproduction requires payment
Cross-border judgments can rewrite rules more than technology itself