Intellectual property in the age of generative AI: deciphering the issue
Increasingly important interactions
The rise of generative artificial intelligence and the worldwide emergence of foundation models trained using self-supervised learning on increasingly vast volumes of data are bringing the interaction between artificial intelligence and intellectual property to the forefront of the debate. The use of content (text, images, sound, etc.) protected by intellectual property law (copyright in particular) for training purposes raises the question of the adequacy of the protection afforded to the holders of the rights to this content. The stakes are high, both for our capacity for innovation and our mastery of artificial intelligence, and for the sustainability of our cultural model: it is therefore imperative to reconcile the preservation of creation with the development of rich, high-performance foundation models on European soil.
A renewed European regulatory framework
The exception to copyright for text and data mining, created by the DAMUN directive, allows the use of content protected by IP rights for AI training purposes, while providing an “opt-out” right for the owners of these IP rights. Faced with numerous debates on this system, at the end of 2023 the European Union included a new transparency regime on content used for training purposes in the AI Regulation (AI Act). Europe has thus equipped itself with a robust body of regulations to apprehend these technological developments. Beyond the regulatory framework, many of the difficulties and concerns that may remain regarding this “opt-out” will be resolved more by technical work than by legislative changes: France and Europe must be pioneers in drawing up the international standards that will become tomorrow’s norms.
Beyond the regulatory framework, structuring challenges
More broadly, there is the question of the volume of data available in Europe for training AI models. This aspect should not be overlooked: it has a direct impact on the ability of our companies and startups to develop AI models on European soil, and will determine the level of cultural and linguistic diversity available to tomorrow’s major AI models.
-
-
Position d'influenceAffaires publiquesNotre feuille de route européenne pour l’essor des startups
-
