AI and droit d’auteur, a round table at the Assemblée nationale

Yesterday morning, the French National Assembly's Cultural Affairs Committee organised a round table on the theme of ‘droits d’auteur et intelligence artificielle’, bringing together intellectual property experts and the heads of the main collective rights management organisations to discuss the impact of artificial intelligence on droit d’auteur, the French version of copyright.

Here's what we learned from these discussions:

• The major problem highlighted by the various stakeholders concerns the exception for text and data mining (TDM) introduced by the 2019 directive, which allows, under certain conditions, AI models to train on huge volumes of data, including cultural data. The application of this exception to AI training is complex, as it depends on conditions that are often problematic, such as lawful access to content and the absence of opt-outs. These conditions are called into question by the frequent presence of pirated content and the widespread potential for opt-outs. In addition, compliance with the three-step test remains uncertain.
• The transparency of training data for AI models is an essential first step in finding out about the works that fed them, without compromising business secrecy. Professor Alexandra Bensamoun sums it up this way: ‘We want to know the ingredients, not the recipe’. The debates revealed a consensus on the need for European regulation of AI, in order to establish a market that is sustainable, ethical and competitive.
• Attempts to negotiate with the various artificial intelligence companies were discussed. It was made clear that, in the absence of a contractual agreement, legal action would be taken, which seems to be the most likely route at this stage for asserting the rights of creators.
• A major challenge is also emerging concerning the remuneration for authors of works whose creation has been assisted by AI. Establishing appropriate levels of remuneration for these creations raises important methodological questions. Defining the criteria for assessing the level of AI intervention in the creative process, and setting up a reliable monitoring system to determine the work's eligibility for copyright and the corresponding level of remuneration, is a complex issue that requires in-depth consideration. This issue will be the subject of future publications by Arvester.