Strasbourg. On March 10, 2026, the European Parliament adopted the Voss report entitled “Copyright and generative artificial intelligence – opportunities and challenges”. He thus invites the commission, which has the monopoly on legislative initiative, to complete the European legislative framework so as to guarantee the effectiveness of copyright in the face of the development of generative artificial intelligence (AI).
Let us make it clear from the outset that Parliament’s resolution is in no way binding. This is a political position which occurs in a context of massive deployment of generative AI models capable of producing output (output) texts, images, sounds, videos and lines of computer code from gigantic data used as input (input) for training AI models, this data itself being largely composed of protected works, including texts, images, sounds, videos or lines of code. Rights holders (performing artists, producers of phonograms and videograms, television and radio broadcasters, publishers and online press agencies) have increased the alerts against the massive use of their productions to train AI models after they have been “sucked” onto the Internet, without authorization or financial compensation and therefore most of the time in violation of intellectual property rules.
Reaffirm copyright
The European Parliament has therefore decided to accelerate. It is clearly in favor of creation, considering that maintaining the economic balance of the cultural and creative industries is perfectly compatible with innovation and the development of AI in Europe. Parliament thus intends to promote creation without opposing it to technological innovation, these two sources of value should not be pitted against each other, but articulated with one another and not one in defiance of the other.
On this ridge, Parliament affirms that European Union (EU) and Member State regulations on copyright and related rights must apply uniformly to all AI operators deploying their solutions in the EU, regardless of the jurisdiction where the copyright acts underlying the training of these models take place and where the results produced by the AI systems are used in the EU. Parliament thus considers ultimately that the law of the Member States must be able to apply when the AI system is used within the EU, including when the servers are located outside European territory. The solution is in accordance with the report of December 18, 2025 of the Superior Council of Literary and Artistic Property (CSPLA) which notably concluded that French copyright is intended to apply to content used for training generative AI models as long as they are exploited in France, even if the act of reproduction took place on servers located outside the EU.
Parliament is thus sending a clear political signal: generative artificial intelligence can develop when it uses protected content only if it is authorized, traceable and paid content, as opposed to the current practice of massive, free and non-consented exploitation of creations.
The European Regulation on AI deemed insufficient
This reminder of the necessary respect for copyright is not new, including within the EU. The principle is in fact already laid down in the Regulation on artificial intelligence (AI Act) of June 13, 2024. This legislation is, however, considered insufficient. And for good reason: the text aims to treat AI systems with regard to the risks associated with the development and deployment of these systems throughout their life cycle. It requires operators, private and public, to put in place compliance programs guaranteeing that AI systems placed on the European market are trustworthy, that is to say lawful, ethical and robust. It is therefore not a text on copyright and related rights, for which it is limited solely to recalling their necessary respect and establishing transparency obligations that are not very restrictive in practice.
Guarantee the effectiveness of copyright
Parliament therefore invites the European Commission to go further than the minimum base set by the AI Regulation. The political roadmap that it formulates aims to guarantee the effectiveness of the rights of authors, performing artists, producers of phonograms and videograms, publishers and online press agencies and even television and radio broadcasters. How ? By requiring suppliers of AI models to produce detailed documentation on the protected content used for training in order to precisely identify this content, and, failing that, by posing a rebuttable presumption of use of the content, like the French senatorial bill adopted on April 8, 2026 after the favorable opinion of the Council of State on March 19, 2026. By ensuring that the measures taken by the rights holders for oppose (opt out) to the use of their content are respected. By proposing licensing mechanisms, including collective and by sector, allowing, on the one hand, AI operators to legally access data and, on the other hand, rights holders to benefit from remuneration. And by retroactively compensating the latter to the extent that their content has already been used, at least since the end of 2022, specifies the resolution.
In short, Parliament calls for the establishment at European level of a standardized legislative framework guaranteeing effective sharing of value between those who create and those who technologically innovate. Will it be heard?
