In this blogpost, Michele Evangelista, EBU’s Senior Legal Counsel delves into 4 copyright principles that AI developers should adhere to when using public service media content: respect for the opt-out, ensuring transparency, proper attribution, and fair remuneration. These principles safeguard copyright, foster collaboration, and ensure lawful and ethical use of media content.

Artificial intelligence (AI) has a crucial role to play in shaping society’s future in support of creativity and innovation. AI systems can help public service media reach broader audiences, enhance their content offer, and can be key to enabling them to fulfil their remit of providing society with reliable, trusted and quality information. But if this new technology offers many interesting opportunities, it also raises significant concerns, particularly regarding copyright protection. It is for this reason that we have identified four principles that AI systems must respect to operate in full compliance with copyright and to recognise the value of public media organisations.

1) Respect for the opt-out

When a broadcasting organisation reserves its rights, AI systems must abide by that decision and seek prior authorisation to exploit its content. In the EU, although in principle providers of AI applications are not prevented from using media companies’ content, public service media can decide to exercise exclusive rights by permitting the exploitation of their content only upon a clear authorisation. This choice to exercise exclusive rights, commonly known as the opt-out, must be assessed by public service media based on their general policy, internal evaluation, and business strategy.

2) Transparency

Transparency obligations require providers of AI applications to disclose what copyright-protected content is used to train the system. It permits to verify if copyright-protected content is lawfully fed into the training data and thereby enables broadcasters to enforce their rights if their content was exploited without the required authorisation. Such openness is the condition for creating a sound collaboration between media organisations and AI systems in full compliance with the principles of copyright.

3) Attribution

AI applications must credit public service media when the output originates from their content. Public service broadcasters are repositories of trusted and quality information that are particularly valuable for AI systems. Since generative AI systems rely on huge amounts of data from many different sources, a distinction must be made between the data originating from public media companies and data originating from other, sometimes less reliable, sources. Crediting public service media guarantees users that the output received by the AI system is trustworthy and helps maintaining the longstanding bond of trust between PSM and society in Europe. This is the reason why AI tools should be designed to ensure proper attribution of the sources used to generate AI outputs.

4) Fair Remuneration

AI systems must pay fair remuneration to public service media that require consent for the use of their content. Remuneration must reflect the value of the content provided by the public service media, which is universally recognized as trusted and quality information, and be proportionate to the proceeds obtained by its exploitation. Since the quality of the service provided by any AI system is proportional to the quality of the data used, the owners of such content should thus join in the revenues hereby generated.

Copyright is a crucial area for AI developers that want to work together with trustworthy content creators, such as public service media. If the above principles can be respected across the board, we face a bright future where AI can support the dissemination of quality information and entertainment.

Are you interested to learn more about the EBU’s work on AI? Sign up for the AI Watch newsletter and visit our AI hub for all of the activities that we are doing to embrace the AI wave.