🔎 The new French bill passed by the Senate on the presumption of exploitation of cultural content by AI providers: a means of combating the lack of transparency surrounding generative AI training

In response to the structural opacity of artificial intelligence providers regarding their training datasets and data supply chains, the French Senate adopted, on 8 April 2026, a bill introducing a rebuttable presumption of exploitation of cultural content by AI systems.

➡️ Direct legal consequences of the opacity of AI systems (AIS):

• the inability of rights holders to prove the use of their works,

• the paralysis of the right to an effective remedy,

• a profound imbalance in the balance of power between creators and AIS professionals.

⚖️ The new proposal introduces the following paragraph into Article L. 331-4-1 of the Intellectual Property Code: “Unless proven otherwise, a work protected by copyright or a related right, within the meaning of this Code, shall be presumed to have been exploited by the artificial intelligence system, provided that evidence relating to the development or deployment of that system or to the result generated by it makes such exploitation plausible.”

👉 Principle: where there is evidence making the use of content protected by an AIS (at the training stage or in the generated result) plausible, exploitation is presumed.

This is a simple (rebuttable) presumption, meaning that the AI provider may provide evidence to the contrary.

🎯 The aim is to shift the burden of proof to those who have technical control over the data and models, rather than to copyright holders, who currently face a near-impossible burden of proof.

The objectives pursued:

•    To restore the effectiveness of rights (intellectual property rights, the right to an effective remedy provided for in Articles 17 and 47 of the Charter of Fundamental Rights of the EU)

•    To go beyond the current mechanisms, which are largely inadequate:

🔹 the Text and Data Mining (TDM) exception, conceived before the rise of generative AI, which is legally fragile and economically unbalanced

🔹 the opt-out mechanism, which is largely ineffective in practice as it provides for no oversight and in which case law – at least in Germany – is becoming entangled.

•    Create a strong incentive for contractual arrangements (licences), rather than widespread litigation

🇪🇺 And at European level?

•    The AI Regulation (AI Regulation) imposes a transparency obligation, but in a form currently deemed insufficiently stringent by rights holders.

•    The French presumption falls within the framework of Member States’ procedural autonomy and the principle of the effectiveness of EU law.

•    The principle of the reversal of the burden of proof is beginning to gain traction and could, in the long term, inspire an evolution of the IPRED Directive (Directive 2004/48/EC of 29 April 2004 on the enforcement of intellectual property rights).

In summary, the presumption of exploitation does not stifle innovation. On the contrary, it seeks to reconcile AI, creativity and legal certainty by rebalancing a system currently based on the opacity of usage.

Image : Pixabay