EWC Statement on Artificial “Intelligence” and the AI Act proposal

The following is the first statement of EWC on AI. It was released on 24 April 2023. It has a good impact in decision makers and our members, who distributed it in their social media and websites. It is the foundation for the many interventions and papers that EWC is developing since then.

Writers of the book sector call for an ethical and legally effective approach to so-called artificial “intelligence” in the AI Act proposal

  • Reform the text-and-data-mining exception

  • Consent requirement instead of dysfunctional opt-out

  • Transparency obligation in the proposal for an EU AI Act (AIA)

  • Independent legal framework on machine output outside copyright legislation

“Any sufficiently advanced technology is indistinguishable from magic.”

––Sir Arthur C. Clarke, Physicist, Science Fiction Writer.

The software and products sold by AI companies today are the result of the non-consensual and non-remunerated use of millions of copyrighted images, private photos, texts, and the work of human individuals, of writers, of citizens of Europe, mined out over a decade.

This exploitation of our work and creative, inventive intelligence, with the aim of establishing competing products, blatantly violates human rights. It damages the sources of the entire European book and cultural arts market.

We are at the dawn of a crisis that reaches far beyond writers: this deep rupture will affect all kinds of jobs and professions, of education and study, of science and media. The rules for dealing with products of so-called AI ethically, economically, legally and in terms of labour policy integrity, however, are man-made: and we demand that they finally be taken.

The European Writers’ Council (EWC), as largest federation in the world championing the rights of solely writers of the book sector since 1977, and representing 46 professional organisations from 31 countries, urgently calls for far-reaching regulation of AI systems and data uses within an ethical, copyright, economic, human rights and privacy rights perspective, with the human originators at heart, as follows:

I – On the input of AI software: A regulated, remunerated, and transparent handling of data and protected works
  1. Reform the existing TDM exception swiftly with a remuneration-requiring design in terms of collective management organisations (CMOs);
  2. Make voluntary opt-in management applicable for each writer. Implement the legally prescribed and currently dysfunctional “machine-readable opt-out” in an enforceable and applicable manner during the transitional period;
  3. Implement licensing obligations for developers of AI products to conclude collective and individual licences with originators;
  4. An explicit obligation of proof and transparency regarding the works used as training material must lie with the developers of AI and, accordingly, with the companies;
  5. In the case of the use of so-called clone voices, which have been used by living persons and audio book narrators without consent, the explicit consent of the human narrators must be obtained, and they must receive an appropriate share of the proceeds.
II – On the output of AI: Implement a Transparency Obligation in the AI Act and regulate AI products outside already existing copyright legislations
  1. A labelling obligation for press, book and text works produced by means of AI, including translations, as well as audio works such as audio books, radio plays, etc., shall be introduced. To further protect freedom of speech, labelling of machine output is absolutely necessary to protect and recognise human achievements. It is important to ensure the traceability of products to also protect consumers who have the right to know what the quality and provenance of the product is and who is responsible for possible damages.
  2. Proposal on an European AI Act (AIA) 2021/0106(COD) – COM (2021) 206: So far, no labelling obligation for AI products in the cultural fields has been provided for, but, in Chapter 5, Title IV, Article 52 (3), an exception for creative or cultural-similar output has been created within the several amendments, esp. 2275 or 2278. Under no circumstances should a non-labelling of any creative-like output by AI software be approved in the Trilogue or the Council; it is a fundamental right of the reader and every cultural consumer to know who wrote or what machine compiled the book.
  3. We support the approach of making not only the information of origins available transparent, but equally transparent when cultural and book works are curated, recommended, or selected for individual purchase by algorithms or otherwise automated systems. The reception and visibility of European book and cultural works should not depend on non-European tools.
  4. Publishers have the obligation to inform authors whether their texts have been (totally or significantly) machine translated. Similarly, we stress the importance of publishers continuing to rely on the skills of human power for editing, cover design and manuscript selection.
  5. AI products should not be protected under authors’ rights and copyright laws; a separate legislation inspired by Directive 96/9/EC of 11 March 1996should be in reflexion for machine made output and products. Under no circumstance a machine shall be considered as an author. Robotic engines compile, mash-up, and compute existing data: they don’t create original art, but products.

The EWC (European Writers’ Council) represents 46 writers’ and literary translators’ associations from 31 countries. EWC’s members comprise over 160.000 professional authors in the text and book sector.

Web: https://europeanwriterscouncil.eu / EU Transparency Register ID: 56788289570-24

Background Information: AI is not intelligent. It’s a mash-up

“There Is No A.I.  There are ways of controlling the new technology—but first we have to stop mythologizing it.” ——Jaron Lanier, Computer Scientist, Author.

For years, the professional works of the over one million book writers of all genres in the European sector have been harvested without consent and without remuneration to develop competing software products of so-called artificial “intelligence”.

Software and machine word processing, often mistakenly referred to as artificial “intelligence”, has nothing to do with the kind of cognitive, emotional, and creative intelligence on which all original works of art, culture and literature are based. There is nothing fundamentally intelligent about the process; automatic writing or machine translations can neither correct the output, nor read, nor understand what they are outputting – and cannot provide any new inventions, thoughts, or findings. The simulation of a conscious entity is a cheap trick, not magic; and at the same time, the cheap hype costs priceless values: it damages copyrights, personal rights, data protection rights, human rights; it restricts linguistic diversity as machines are trained mostly in English, produces bias and prejudices, disinformation, remuneration cuts, broaden the breach of inequality among countries and persons, and are a devaluation of the unique originality of works crafted by human intellect and skill.

In publishing houses, manuscripts are more and more reviewed by software for their “sales potential”, in online retailing the algorithm has replaced personal recommendation, keyword tools summarise book works and generate marketing texts from them. Translators are less often hired for translations; instead, post-editors are assigned to revise machine translations. In the children’s book and novella sector, self-publishing service providers like Amazon are flooded with automatically produced picture books and ChatGPT texts. Human voices of audio book narrators are cloned and used without consent. For example, the voice of actress Emma Watson (Harry Potter) was synthetically copied and had her read out Adolf Hitler’s “Mein Kampf”.

Worldwide, companies such as Oracle, Alibaba, Google, Microsoft, OpenAI, Nvidia or Apple and Amazon are working since a decade on text generators, machine translation and synthetic dubbing of text works into audio books. The data sets for the “training” of translation and self-writing software are based on public domain text works accessible on the internet, and on copyrighted books from the years 2013-2021 and younger, which come from piracy sites. [1]

The regulation on text and data mining (TDM), which has been in force since 7.6.2021 through the Articles 3 and 4 of the Directive on Copyright in the Digital Single Market (EU) 2019/790, also makes it shamefully easy for non-European companies to mine training material: the “machine-readable opt-out” legally possible for online published works, the so-called licence or usage reservation instead of a consent requirement, is not applied anywhere, as there is neither a technical nor contractual arrangement for this. In addition, the transposition of these articles in national frameworks, excludes any compensation for this usage.

We call this: theft, parasitism, and a short-sighted policy that allows non-European companies to benefit from the output of European intellectual sources. However, the drain of intellectual capital does not only have a direct impact on writers themselves: already, non-cultural companies are replacing text, visual or interpreting workers with automated software.

At the same time, the allocation of responsibility for e.g., violations of personal rights or disinformation by machine-generated products is unclear.

The over 160.000 individual writers of the book sector represented by the European Writers‘ Council (EWC) and it’s 46 member organisations, deserve a strong, clear and based on European values response from all legislative and economic decision makers involved in the cultural sector. As the current hesitancy of European policies is not solely a threat to written culture but democracy and the future of the values of human achievements, we plead for bold action.

Demystify AI. Put the human being and European citizen at the centre of all efforts. Only in this way can the achievements of the technology be regulated in a social, ethical, and dignified way.

Brussels, 24 April 2023

The European Writers’ Council (EWC) Board.

Nina George               Miguel Ángel Serrano             Maïa Bensimon

President                      Vice President                          Vice President

Germany                      Spain                                            France

Eystein Hanssen          Alena Makouskaya                 Markku Loytönen

Board Member             Board Member                         Boad Member

Norway                            Belarus                                         Finland

[1]  https://aicopyright.substack.com/p/the-books-used-to-train-llms

Further ressources on this page: https://europeanwriterscouncil.eu/artificial-intelligence/

Proposal for a Regulation laying down harmonised rules on artificial intelligence {SEC(2021) 167 final} – {SWD(2021) 84 final} – {SWD(2021) 85 final}