In this trilingual address at EuroVisioni, Roberto Viola, Director General of DG CONNECT, outlines the European Commission’s strategic vision for the audiovisual industry. Beginning with Artificial Intelligence, he highlights its production benefits while emphasizing the AI Act’s role in managing risks like deepfakes and copyright infringement through transparency obligations and a new code of conduct signed by major global players. Shifting to Italian, Viola discusses public service media, stressing that the Media Freedom Act (EMFA) now legally safeguards their independence and stable funding, which is crucial for the European information ecosystem. He argues that public broadcasters must modernize via digital platforms and AI to remain relevant to audiences who increasingly consume news via social media. Furthermore, the speech addresses audience measurement, where the EMFA introduces standards of transparency to correct market imbalances caused by large internet platforms. Finally, speaking in French, Viola presents “Agora EU,” a comprehensive funding program including a €3 billion “Media Plus” envelope, and calls for a modernized Audiovisual Media Services Directive to ensure that Europe remains a global reference in terms of creativity, cultural diversity, and democracy.
___________
It’s a great pleasure to join the edition of EuroVisioni this year, unfortunately from remote. I think EuroVisioni offers always an opportunity for discussing and reflecting on key topics that shape the media and audiovisual industry.
I would thus first like to thank the organizers because this year the program is absolutely very rich, and all the panels are really touching on burning questions and key elements that will shape the future of audiovisual industry. Let me just mention AI, public service media, audience measurement, and the future of our cultural and creative industry.
Balancing Innovation and Regulation in AI
Let me start with AI. AI is, of course, at the center of the focus of the action of the European Commission, and AI also promises which are very important for the audiovisual media sector. AI can support automation of visual effects, enhance camera-based video production, reduce the cost of production, and even artificially rejuvenate actors. I would say the plusses of utilizing AI are much more than the minuses. However, at the same time, we are aware that the introduction of generative AI can raise concern around important issues for the sector, such as copyright, authorship, and content authenticity. The use of existing audiovisual works to train AI models, it’s useful, but at the same time the creators must be respected. Also, it’s important to use AI to create, but deepfake should be combatted.
Regarding copyright, the EU AI Act includes two provisions to aim to facilitate the enforcement of copyright rules while increasing transparency on the use of protected works in the training of AI models. As you know, the AI Act is the first world comprehensive legislation on AI, and we have been concentrating our attention in the first part of application of this law to the so-called General Purpose AI models, the large models, and how the provider of this model would respect fundamental issues such as safety of the algorithm, respect of fundamental rights, and of course copyright law.
Copyright Transparency and the Code of Conduct
According to the AI Act, providers have to draw up and publish a sufficiently detailed summary about the content which has been used to train the models. And this obligation, it’s specified by the European Commission AI Office, determining which kind of disclosure needs to be done. For the first part of the applications the AI Act, the first step is a code of conduct. After one year of work, the code of conduct on large AI model has been signed on the 10th of July 25, and 27 major EU and US players have signed this code of conduct. This is an unprecedented result; all the important players in the world have signed the code of conduct on AI. Now, the final version of this code of conduct has a very robust copyright chapter that I think reflects a balanced compromise between the key concerns expressed by right holders, but also the concerns of providers, especially startups, to have legal certainty when it comes to use copyrighted works. The added value of this chapter is that it specifies a set of clear workable measures, consistent with the EU copyright law, that providers commit to implement. For example, proportionate and appropriate measures to prevent copyright infringing outputs.
As I was mentioning, another part of the implementation of the AI Act when it comes to copyright was to publish from the European Commission AI Office the template which addresses the AI provider in terms of disclosing how they have trained the model. The template allows now right holders to know which key websites, and which key resources the providers have utilized to train the model, and to check this against their rights in terms of having exercised the opt-out which is foreseen by the copyright law, or the respect of the agreements they signed with the providers.
Regarding the authenticity of content and deep fakes, the EU AI Act sets out transparency obligation for providers and deployers of generative and interactive AI systems. Such transparency obligations support distinguishing AI content from human content and intend to address the risk of deception and misinformation. One obligation that is particularly relevant for actors of the audiovisual sector, especially professional users, is the obligation to signal when a content has been generated through AI means. Now, this obligation is well balanced with another type of AI generation, which is the generation of creative AI contents in the context of movie-making, audiovisual production. So, the AI Act tries to reach an equilibrium between, I mean, what are the obligation of transparency, for instance, in information programming, and what are the rights, the fundamental rights of freedom of expression of authors and creators when using AI in the context of artistic creation. This balance is fundamental and this balance must be kept at European level because at national level the utilization of different technique, different ideas would create only confusion.
In order for these two fundamental elements of the AI Act to be effective, the next step is a code of conduct that will allow all the participants of the audiovisual sector and the other interested stakeholder to draw the measures they are going to utilize when offering AI content to the viewers. We will now start the work of the code of conduct. We have received 200 stakeholder applications, and many coming from the audiovisual sector, to participate to this new code of conduct. We will start the work basically this month, and we expect to publish the final version by June 26. We, as we have done for the other codes of conduct on AI, we will publish successive drafts, and the first draft of this code of conduct will be published by the end of the year, and I invite you all to provide us feedback on the work so that your valuable feedback can be utilized in the final version of the code.