End the "pillage" of authors by AI: the law proposal validated by the Council of State

Also read: After the failure of negotiations between tech and rights holders, soon a new law to protect creators?

A text neither contrary to the Constitution nor contrary to European law, according to the highest administrative court

Today, it's the opposite: it's up to the authors to prove their work was used. However, it is « almost impossible » for creators to prove that such content was used to train an AI tool. AI developers do not publish an exhaustive list of their training data. The system indeed relies on « full transparency from AI providers », which does not exist, noted the authors of the bill in their rationale.

The Intellectual Property Code would thus state that:

« unless proven otherwise, the work protected by copyright or by a related right, in the sense of this code, is presumed to have been exploited by the artificial intelligence system, provided that an indication relating to the development or deployment of this system or to the result generated by it makes this exploitation likely ».

Also read: Will AI companies have to pay damages to authors? The complex issue should be resolved in 2026

But according to the tech lobby, this new rule would penalize French AI companies and lead to significant loss of competitiveness, arguments supported by the French startup Mistral AI, which developed Le Chat, and France Digitale, which represents the interests of startups. These companies questioned the compatibility of the text with European law: all elements that led the President of the Senate to refer the matter to the Council of State for an opinion on February 19th.

Also read: Why this vote by the European Parliament could change the situation for authors and artists facing AI

The text does not change European rules, but merely provides for a specific evidentiary regime, according to the Council of State

In detail, the highest administrative court estimates, in a consultative opinion published on its website, that the proposed law « is limited to providing for a specific evidentiary regime for infringements of copyright and related rights, without changing the rules and principles that define these infringements ». However, the 27 national legislatures are well competent to set these « procedural modalities (…) of legal remedies aimed at ensuring the protection of rights ». Result: « the national legislature is (well) competent to establish such a trial period ».

In addition, « the bill aims to ensure the effective protection of copyright and related rights, which fall under property law (…) while seeking to establish equal footing in the course of the trial allowing the defense of such rights », writes the highest administrative court.

The Council of State also estimates that the bill does not need to be notified to the European Commission. This latter body finally refutes an argument put forward by the tech industry, according to which « the index » triggering the presumption would be too vague. The index could be translated into « responses generated by the AI system, technical assessments, scientific articles, but also public revelations made by AI system providers » or even « several factual elements », continues the highest administrative court. 

In parallel, Mistral advocated, on March 20th in an article in the Financial Times, for the establishment of a different system. Rather than a presumption that would now weigh on AI companies, these companies should contribute to the financing of a fund intended to compensate authors whose content has been used without authorization. The Senate bill proposal will now be examined by the upper house in a public session on the upcoming April 8th.

Also read: Copyright: From August 2nd, AI will have to properly disclose their sources in Europe

Also read: Salary halved, fewer contracts, loss of meaning... Translators hit hard by generative AI

👉 Follow tech news in real time: add 01net to your Google sources, and subscribe to our WhatsApp channel.