SAA - Society of Audiovisual Authors

There needs to be a clearer differentiation between data and copyright-protected works

We understand that the Commission is pushing for the development of AI in the EU, with a view of making the continent an attractive place for investing in this technology. However, discussions on the data needed for training AI and therefore the creation of a European data space where data is seamlessly shared, is often too generalist in terminology and does not appropriately take into account that some of that data is in reality works of authorship, protected by copyright.

Lately, there has been a tendency to refer to copyrighted works as ‘high-quality data’, indicating that they are different from other types of data that might not be as valuable for crisp development of AI, in terms of its reasoning, processing of information, etc. Yet, the classification as data remains.

Works of authorship are not mere data. This classification reduces the importance of copyrighted works to just another piece of data in the mix of millions of data used to train and develop AI, as if these works were not the expression of authors and the building block of European culture. Protected works are not an automatic registration of access to an online service or tracking of weather conditions.

Additionally, treating copyright-protected works as mere data devalues human labour. Works of authorship are the result of hours of working time spent in redefining the author’s craft and bringing the idea to life. This effort cannot be devalued to the point of becoming another line in the millions of lines that make up a dataset.

Therefore, in the discussion concerning data and the next steps in building a Data Union Strategy, it is of fundamental importance that clear distinctions are made in addressing the data and the copyright protected works that are being used for technological development. This not only recognises the value of human authorship but also helps in raising awareness among the market operators, including the SMEs, start-ups and all entrepreneurs that want to make data the centre of their trade.   

Copyright protection still applies, authorisation is needed, and remuneration is due

Regardless of how protected works are called in the data conversation, there is an even more basic issue that is not addressed enough in the present consultation: copyright protection applies to works, authorisation is needed, and remuneration is due. Copyright protection is not an hinderance to data sharing.

Any data sharing situation shall respect the copyright protection of works, as any use of works shall be authorised by the rightsholders or their representatives, notably CMOs. Licensing is indeed the most secure and certain way to use works, eliminating the risk of copyright infringement for unauthorised commercialisation.

Speaking specifically about AI development, it is important that the discussions on the EU data strategy take into account the current framework of uncertainty regarding the use of protected works. While the TDM exception is considered to be the way AI companies can develop their models without prior authorisation, issues regarding the conditions of applicability of the exception such as lawful access, means of opting-out and even the very application of the exception itself in the AI context are still strongly debated and part of ongoing studies and reports commissioned by EU institutions and agencies.

This means that the Data Union Strategy must consider the developments in the AI and copyright framework and exclude any action that has the potential of creating even more uncertainty and risks for businesses dealing with data (either shared between Member States or with international partners) due to contradictions in purpose, principle and provisions.

In any Data Union Strategy that works for the European market, protection of intellectual property rights, including copyright, shall be a priority and a facilitator to ensure the prosperity of all operators participating in the market. This includes authors and other rightsholders, who should not bear the cost of technological development and should instead be actively included in value sharing: the Data Union Strategy shall promote a healthy licensing market, where both audiovisual authors and technology innovators can thrive.

Any data sharing activity that involves works of authorship shall ensure that authorisation for the use of protected works and remuneration to the authors of such works are duly provided. On one side, this ensures that AI companies can develop their models with original works that are embedded with diversity and European values by design. On the other side, this approach will maintain the richness of expressions that characterise the EU, created by properly remunerated European authors.