Gen AI and copyright: what role should European protected data play in the AI era?

With the deployment of AI, new fundamental questions arise at the intersection of creativity, competitiveness, sovereignty and, in France, cultural exception. Can generative AIs be trained on content protected by copyright? Is it possible to find common ground between rights holders and AI model developers? Could potential disagreements jeopardize the use of European data in future generations of AI tools and applications, potentially strengthening the influence of non-European content in AI used by Europeans?

These are the questions we aim to address in this study. However, the answers are not simple.

On the one hand, for the sake of European cultural influence, the plurality of information, and the representation of our democratic values, the training datasets for AI models should reflect our linguistic diversity and the richness of our intellectual and artistic produc- tion. Access to European cultural data should therefore be guaranteed to all model providers, especially European ones.

On the other hand, the cultural sector is concerned about the impact of generative AI on their business models and the future of their professions.

So, how can we find a balance between these two perspectives? Are they irreconcilable?

We are convinced that playing the opposition between innovation and culture is a futile battle that will benefit neither European businesses nor content creators. The proposals presented in our study, will require in-depth discussions among various stakeholders. France Digitale is ready to take part in this dialogue along with the European innovation ecosystem.