As part of the Belgian presidency of the Council of the European Union (EU), the FPS Economy invited experts and organisations to a two half-day copyright conference that took place in Namur on April 8 in the afternoon and April 9 in the morning. The theme of the conference was ”The Action of the EU and its Member States in Favor of Fair Remuneration for Authors, Performers and Creative Industries in the Digital Content Landscape”. The EWC’s Commissioner for Political Affairs, Ms. Nina George, took part as Speaker on the Panel on Artificial Intelligence and Copyright.

Together with European and national copyright experts, the Copyright Conferene focused on:

• how the DSM Directive was implemented across EU Member States
• how to improve the remuneration of authors, publishers and performers in today’s online environment

In this context, several experts presentations were held, offering potential solutions to shared challenges, followed by panel discussions with relevant stakeholders. In addition, the Conference looked in future tasks for European copyright, with topics such as artificial intelligence and contractual buyout clauses.  Among the speakers: Daren Tang, Director General of WIPO, Roberto Viola, Director General of DG CNECT, Frank Gotzen, President of ALAI and Professor, Emmanuelle Du Chalard, Head of the Copyright Unit of the European Commission, Iban Garcia Del Blanco, MEP, Giuseppe Abbamonte, Director for Media Policy at the European Commission.

To the full program

To the presentations

Artificial Intelligence and Copyright: The EWC stressed the need for remedies of past copyright infringement to built the existing generative AI software

In the discussion panel (9/4) on the use by AI of protected works (not covered by the TDM exception, or when rightholders make use of the opt-out faculty) and the remuneration of rightholders on the one hand, and the protection of the result produced by AI on the other hand, moderator Bernt Hugenholtz discussed with:

• • Nevena Kostova, Google
  • Paul Keller, Open Future
  • Agnieszka Horak de Sousa, IFPI
  • Nina George, European Writers Council

 

 

 

 

The exception of Text and Data Mining is highly in question and needs a formal clarification, said Nina George. Until then, it’s an artifical umbrella with loop holes

“As we know, the interpretation of how extensive the TDM exception 4 of Directive 2019/790 applies is far from settled. It is technically and legally highly questionable whether it covers uses that clearly go beyond TDM, such as the development of large language models, AI algorithms or generative informatics. TDM can be one of several procedures before AI processing, but this does not solve the problem of copyright infringement by generative AI developers.

Until this is formally clarified, the opt-out can only act as a maybe well-intentioned but poorly thought-through protective shield – an artifical umbrella with loopholes:

• First, the entire burden of action is placed on authors and rights holders for self-defense.
  The more time you spend on the action steps to be taken, the more horrendous the investments of authors and other rightsholders become: Administrative and legal costs for adapting new and existing contracts. Technical upgrading of electronic files and the digital distribution chains all the way to the websites of authors, publishers and online booksellers: everyone has to invest money to establish the technical requirements and, above all, to adapt the protection shields to even more crawlers every year. And we are not talking about pennies here.
• Still, open questions need to be clarified, such as: how to deal with published works? Half a million titles are published in Europe every year, and 13.8 million books are in circulation in Europe today. Should we introduce a date limit? What about books published in Summer 2021 and that have already been used? What if works by humans are fed into existing generative systems because they don’t read like a machine, but need a different indication? Can an opt-out be applied to existing books, and what do we do with the fact that established generative AI systems used books from piracy sites?
• Second, there are technical terms not clarified sufficiently when it comes for example to books in all formats, electronic epubs, mobi, pdf, but also print or audio books: What is machine-readable? Texts, codes, barcodes, unicodes, optical symboles, chiffres and towards which machines? The Metadata and Onix or robots.txt are far from standard yet. Where can other formally declared opt-outs be applied – imprints, terms & conditions, banner on a website? How can it be secured that this is respected, also in cases when a print book from 2018 is digitised and offered on piracy sites, or sublicensed ignoring a rights reservation within the imprint?
• Third: how can we guarantee that crawlers, Corpora-Builders and AI companies will respect opt-outs, how can we document the proof for lawfully access to works, as laid down in the Directive – as there are no contact persons and the current refusal of companies to be transparent is not exactly something to be optimistic.”

 

 

If you really want to have legal certainity when developing generative informatics based on copyrighted works, you should not rely on the TDM exception Art 4, but license them.–Nina George, EWC’s Commissioner for Political Affairs