AI opt-out system won't help authors' remuneration
On 7 March, I was invited by the Hellenic Copyright Organisation to a conference they organised for their 30th anniversary in Athens. Many prominent copyright experts gave valuable presentations on copyright and AI, while I participated in a round table on Copyright and Collective Management in the Age of AI.
The conference was moderated by Prof. Bernt Hugenholtz. Alongside my colleagues from GESAC (Adriana Moscoso del Prado), AEPO-ARTIS (Ioan Kaes) and IFRRO (Anthi Akritidou), I explained the challenges CMOs face in the age of AI.
SAA’s members have expressed interest in licensing the use of the works of the audiovisual authors they represent for the purpose of training AI models, but they face many obstacles along the way.
In a number of countries, audiovisual authors’ CMOs’ activities are based on a licence prescribed by law and delivered by the national authorities to manage specific rights and fields of exploitation. These laws and licences need to be adapted to include the use of works by AI training models.
In countries where such an authorisation is not necessary for a CMO to develop a new licensing activity, CMOs may need to adapt the mandate of their author members as the training of AI models is a new field of exploitation. CMOs may have to amend their representation agreement in line with their bylaws.
Where CMOs have such an authorisation or mandate, they have usually expressed a reservation of rights according to Art 4 DSM and contacted AI companies to initiate the negotiation of a licence. However, all this has so far been in vain as AI model developers are not interested in entering licensing negotiations.
This situation is the result of regulation and market development that are not conducive to any licensing negotiation. It is important to realise that the AI Act is not sufficient to develop a licensing market in which CMOs would play a role.
The reservation of rights is insufficient to trigger a licensing incentive for AI companies.
Rights reservation is a very difficult requirement to handle for individual authors. We are not aware of any technology/tool that is convenient for authors as they are not the distributors of their works, which have usually been licensed to many operators, making their works available on different types of media/platforms. When CMOs have expressed this reservation of rights on behalf of their authors, international AI companies did not even answer their letters.
The problem lies with the misinterpretation of the impact of the opt-out and the reservation of rights, which stems from the EU Commission’s belief that the reservation of rights is the way for rightsholders to regain control of their works. AI companies, however, see opt-out as a prohibition to use the works, so there is no incentive for licensing in this context.
Improving the opt-out system will not help authors.
I do not believe in solutions like an EU-wide repository of opt-outs that would put the burden of proof on the authors’ shoulders and would raise an additional question on formalities to benefit from copyright protection, risking positioning the EU de facto outside of the Berne Convention.
It would create more problems than it can solve: who should handle such a repository and pay for it, who would be responsible to input the data, etc. All this just for facilitating the life of AI companies who do not want to pay for the use of the works and will not be further incentivised with such a repository. In the end, improving the functioning of the opt-out system will not help authors and their CMOs in developing agreements with AI companies and be remunerated for the mass-uses of their works.
New measures are urgently needed to develop collective licensing for authors’ remuneration.
We believe action at the EU and Member States level is needed to develop collective licensing that will deliver remuneration to authors. This is backed by the Member States in their contributions to the Hungarian Presidency questionnaire on AI and copyright. As the Presidency summary states: “A significant number of Member States were of the view that the mass scale of AI-related copyright uses and the practical challenges of the rightholders in monitoring them both underline the importance of collective management in this area”.
The European Commission’s Executive Vice-President for Tech Sovereignty, Security and Democracy, Henna Virkkunen, also stated in her written answer to the European Parliament before her hearing: “Supporting the development of a well-functioning market for licensing copyright-protected works for AI purposes, with balanced terms for different actors involved, is essential for ensuring appropriate remuneration for creators and creative industries while ensuring competitive access to high quality data for AI providers, particularly smaller EU providers.”
A new approach should therefore acknowledge the limited application of the TDM exception in the case of generative AI and reaffirm the principles of authorisation and remuneration for the use of authors’ works for AI purposes. This is the only way to force AI companies to seek licences.
This is also important to ensure the diversity of the training data used. High quality data means human authored works. So, if we want these tools and models to reflect European cultures, languages and creativity we need to ensure that AI companies are obliged to enter collective licensing with CMOs representing large European repertoires in the different creative fields.
CMOs are well equipped to handle this, they already manage collective agreements with platforms such as YouTube and handle millions of works and data. There is no other organisation better placed to provide such comprehensive licencing agreements.
SAA Secretary General
The recording of the conference is available here.