Joint letter in response to the dialogue with the Audiovisual Sector on Copyright & AI
Brussels, 30 November 2023
Dear Mr. Abbamonte,
First of all, we would like to thank you for organising the meeting on Copyright and AI on 24 October 2023, in the context of the audiovisual stakeholders’ dialogue.
As our organisations emphasized a number of times during the informal consultation on the future challenges of the audiovisual sector, and in light of the increasing significance of AI in the production of new content, we believe that the European Commission should refrain from presuming that the existing legal framework, notably encompassing copyright and data protection rules, is adequately equipped for this radically new technological landscape.
While we acknowledge the hesitance to revisit these instruments, given the substantial efforts invested in their approval, we argue that this is not a valid reason to maintain their suitability. It is imperative to scrutinize their applicability to the realm of AI, and in particular generative AI, clarify their scope and purpose as needed, and promptly rectify any identified shortcomings.
Contending that the fundamental interests that EU copyright and data protection regulations aim to safeguard can still be preserved in a dramatically new technological landscape is misguided. Specifically, when AI engages with the intellectual property of content rights owners, such as in the extraction and scraping of copyright-protected content for training, we assert that depending on the Text and Data Mining exception for commercial purposes (as outlined in Article 4 of the CDSM directive) is a debatable shortcut. It represents an unacceptable prioritization of corporate interests who exploit our works over the creators we represent. This technology is trained on colossal amounts of data, a substantial portion of which comprises the work contributed by our members and for which they are granted IP protection at Union level.
This exception – which was not subject to an impact assessment – was conceived, negotiated and adopted at a time when generative AI had not yet been made available to the general public. AI and generative AI are nowhere explicitly mentioned in the EU 2019 copyright act. Assuming that the text and data mining exceptions should now also extend to this technology would be a quantum leap and a dangerous circumvention of EU democracy since neither the European Parliament nor the Council ever expressed a position related to generative AI in 2019. We also contend that the vague reference to the “development of new applications or technologies” in Recital 18 of that directive could not reasonably be intended to anticipate a technology of this magnitude.
Expanding the Text and Data Mining (TDM) exception to encompass generative AI and content reproduction for the purpose of training this intelligence and generating new content, in our perspective, does not meet any of the conditions enshrined in the 3-step test. This crucial test asserts that exceptions, including for acts of reproduction undertaken for text and data mining purposes, can only be applied “in certain special cases that do not conflict with the normal exploitation of the works or other subject matter and do not unreasonably prejudice the legitimate interests of rightholders”. Text and data mining, as defined in the Directive, refers to an “automated analytical technique aimed at analysing text and data in digital form in order to generate information which includes but is not limited to patterns, trends, and correlations”. The stated purpose of this technique is to acquire “new knowledge” and facilitate the discovery of “new trends,” thereby benefiting the research community and fostering innovation, very far from the commercial use of generative AI today. Without compensation granted to rights owners and considering that AI can generate novel content unfairly competing with the works it is trained on, we consider that this use may indeed conflict with the normal exploitation of the work and unreasonably prejudice their legitimate interests. Generative AI exploits protected content in opacity and with impunity, not for development purposes but rather for operational use in generating new content.
Even if the scope of Article 4 of the CDSM directive were considered inclusive of generative AI, which we refute, the ability to reserve rights is nothing more than an illusion for the members we represent. Their ability to reserve rights for making reproductions and extractions for text and data mining is in fact significantly impeded by prevailing contractual practices. This means that, in the first place, they may not have the opportunity to hold on to these rights, while their contractual counterparts – provided they manage to successfully reserve their catalogues, which is not guaranteed – will typically want to retain these rights primarily to maximize the licensing value of the content. Unfortunately, these deals seldom result in benefits for the content creators. Even for self-produced creators who do not engage in contractual relationships with publishers, producers, or distributors, attempting to assert their right of reservation for content made publicly available online in machine-readable format proves to be an impractical task in practice.
We also want to highlight the symbiotic relationship between numerous tech companies engaged in AI development and the scientific ecosystem. Many of these companies actively collaborate with universities and research organizations. Even where content rights holders may exercise their right of retention under Article 4 of the CDSM directive, it is relatively straightforward for AI companies to bypass this obstacle by leveraging data covered by the mandatory exception under Article 3 of the directive. These current practices also obviously prevent licensing opportunities for TDM despite the fact that Article 3 of the CDSM Directive provides that “rightholders should remain able to license the uses of their works or other subject matter falling outside the scope of the mandatory exception (…) for text and data mining for the purposes of scientific research”.
Finally, it is crucial to emphasize that generative AI frequently carries significant implications for moral rights, specifically the right of attribution and the right of integrity, as safeguarded by the Berne Convention, a treaty endorsed by all EU Member States. While there is a lack of harmonization at the EU level regarding moral rights, most EU member states do recognize and uphold these rights as well. In the context of generative AI and deep fakes, preserving these rights is paramount to safeguard the honor and reputation of authors and performers.
For all these reasons, we strongly believe that discussions regarding the future of the EU audiovisual industry must not evade a comprehensive analysis. This appraisal should aim to assess whether the copyright acquis (and, indeed, the general data protection regulation, which we are not specifically addressing in detail here) is suited for a world where AI extensively utilises copyright-protected (as well as biometric) data to generate new content. In this context, creators lack the ability to authorize or prohibit such use, let alone to derive any kind of compensation from such use.
Avoidance is not an option. It is necessary to confront the challenge of constructing a fair and equitable AI, trained in full compliance with copyright and data protection principles. Moreover, it must respect the rights of those who do not wish their personal and non-personal data to be used for this purpose, especially without adequate compensation. We therefore call on you to continue this discussion and engage the Commission in assessing the technological developments and their impact on the audiovisual sector with special attention to the impact on its creators.
CC: Renate Nikolay, Deputy Director General – DG CONNECT
Eleonora Ocello, Member of the Cabinet of Internal Market Commissioner Thierry Breton
Signatories
ECSA (European Composer and Songwriter Alliance)
FERA (Federation of European Screen Directors
FIA (International Federation of Actors)
FSE (Federation of Screenwriters in Europe)
SAA (Society of Audiovisual Authors)
UNI MEI – UNI – Media, Entertainment and Arts