<i>GEMA v. OpenAI</i> : The Munich Regional Court Presents an Exemplary Judgment on a Well Conceived Test Case

Abstract The judgment presented here, GEMA v. OpenAI, has raised considerable attention. It is the first case concerning reproductions in an AI model and uses in the output of a chatbot based thereon. For this carefully prepared test case, GEMA had secured the rights for nine song lyrics, exercised the opt-out, and obtained through simple prompts copies or only slightly modified copies of those lyrics. The Munich Regional Court held that this output clearly proved that reproductions were contained in the model and that the lyrics were reproduced and made available to the public in the output. Since none of the reproductions served an automated analysis as described in the definition of text and data mining, the mere wording excluded the application of the TDM exception; importantly, the Court also stressed the rationale of the TDM exception, which was to allow only a mere analysis, which does not affect the authors’ exploitation interests. The Court also established convincingly that it was the defendant rather than the user of the chatbot who had performed these relevant acts and was thus directly liable, so that the criteria for intermediaries’ indirect liability could not be applied, and OpenAI is liable for injunction, damages and information on the use. It is to be hoped that the higher instance courts will follow this straightforward interpretation and lead to a much-needed clarification of the existing law, while, in parallel, new legislation as proposed by the European Parliament should re-establish meaningful and working authors’ rights for GenAI uses.