Own-Initiative Report on Cultural and Creative Sectors in the Age of AI (2025/2180(INI)), Rapporteur: Hélder Sousa Silva MEP, CULT

Submission by the European Writers’ Council (EWC): Europe’s competitiveness does not lie primarily in simply replicating the same model, albeit with a few more legal safeguards. Europe’s competitiveness lies in the future of what is – for the time being – its third-strongest economic driver: the creative and cultural economies.

We call for a change in Europe’s mindset. We have everything – and much more besides – that so-called generative ‘AI’ cannot achieve in this century: courage and determination, cultural distinctiveness and new ideas, and a political will to shape our future together, which advocates for a sense of ‘us’. Accordingly, political regulations should focus less on how to relate to AI. Regulations should impose more regulation of AI technologies, and offer culture more – enforceable – rights, as well as foster the living and working environments of the individual originator. 

Brussels, 1 September 2026

Dear Rapporteur Hélder Sousa Silva, dear Co-Rapporteurs,  dear Members of the Culture and Education (CULT) Committee:

The European Writers’ Council (EWC) expresses its gratitude for your invitation to give both oral and written feedback prior to the Own-Initiative Report on Cultural and Creative Sectors (CCS) in the Age of AI (2025/2180(INI)), and for inviting us to outline the most important points on behalf of Europe’s 1.2 million book writers and translators at the stakeholder meeting on 2^nd September.

Download the full Submission (pdf, 13 pages)

Summary

Our submissions are placed within Chapters III-VII, adhering to the topic-based structure requested by the rapporteur.

For each aspect, we first set out the conclusions and requirements and supplement these with more detailed background information where appropriate.

I. Preamble: About the EWC and its position and experience with political regulation on AI:

• Writers are the originators of the value chain, yet they are the last to receive support

II. Sovereign wording, confident stance: to know what AI is and what it is not is key to regulating it

• Gaps in the European policy draft Own-Initiative Report should be filled with a technically informed approach and vocabulary.

III. The opportunities and challenges arising from the use of AI in the cultural and creative sectors:

• Machine-generated products have no economic benefit for writers and translators – only for fraudulent actors.

IV. The protection of copyright and related rights, including transparency regarding the use of protected works for AI training and the fair remuneration of creators and other rightsholders:

• The EWC’s remedy calls for the removal of the TDM exception, and future regulation under the ART-principle (Authorisation, Remuneration, Transparency);
• The EWC calls for trust and transparency as enforceable key principles: neither the template of the AI Act nor the transparency obligations are sufficient to let Europe’s CCS survive.

V. The impact of AI on creative professions, working conditions, skills and employment:

• The EWC calls for a comprehensive EU-wide study on the economic harm done by AI products related to: Input; Output; and impact on income, pension schemes, tax, and cultural funding;
• The EWC calls for checks and balances for literacy, and to continue job creation action plans including access to adequate social protection for those affected by AI-related job losses;
• The EWC calls for safeguarding the freedom to say “no” to any AI-adaption of their works.

VI. The preservation of cultural and linguistic diversity and the visibility of European works in the digital environment:

• EWC proposals for targeted funding, exchange and educational programmes;
• The EWC calls for a legislative work on active anti-discrimination in the discoverability of European cultural works, like the AVMS Directive (EU) 2018/1808.

VII. Possible policy and regulatory measures to support responsible innovation and the sustainable development of Europe’s cultural and creative sectors.

 The most important priorities for policy and regulatory measures:

• Call for an EU wide study on the market harm and socio-economic impact of so-called generative AI;
• Seek remedies for the damage caused by copyright infringement;
• Remove the non-applicable TDM exception(s), to allow voluntary and individually negotiable licensing by authors against appropriate and proportionate payments, without posing any further “AI exception” or mandatory licensing upon authors and further rightsholders;
• Encourage Member States to set up mediation entities for authors who wish to say “no” to the use of their works within TDM, GPAI, or generative AI usage;
• Initiate an adjustment of EU competition law, so that authors’ associations are allowed to publicly advise on contracts, clauses, remuneration, and how to deal with unauthorised AI adaptation;
• Set up a human-readable labelling system for AI-fabricated products of all kinds;
• Request the initiation of a legislation on active anti-discrimination for the discoverability of European cultural works;
• A dedicated Directive for the social protection of authors, artists, and cultural professions.

 

FULL VERSION

1. Preamble: About the EWC and its position and experience with political regulation on AI

Writers are the originators of the value chain, yet they are the last to receive support.

As the only federation worldwide representing the interests of authors of the book sector, the European Writers’ Council (EWC) is a non-profit, non-governmental federation constituted by 55 writers’ and literary translators’ associations in 34 European countries, from the EU, the EEA countries Iceland and Norway, as well as Belarus, Switzerland, Montenegro, North Macedonia, and the UK. EWC Members comprise over 260,000 professional authors in the book sector, writing in 37 languages and publishing in all genres (fiction, non-fiction, educational, academical, children & young adult, poetry).

With at least 7.2 million titles stored in the foundation of the most profitable data banks – the existential foundation of all existing automated chat bots – European writers and translators are the most ripped-off independent cultural workers. Our works have been taken without permission for the profit of (mostly) non-European technology companies.

There are currently over 1,000 so-called large language models (“LLM”) in use worldwide, dozens of which have a computing capacity of 23^10 FLOP and a database of several million protected book works. Many of these books were published well before 2021 and the entry into force of the CDSM Directive (EU) 2019/790, but these databases also include books published after 2021, incorporated by the devastating over-interpretation of the TDM exceptions. The creation of these databases is a mass violation of copyright, of authors’ rights, and an unprecedented abuse of people’s creativity, their working hours and their investments.

Writers are never paid for the thousands of hours involved in crafting a manuscript over years, years of personal financial investment and self-exploitation. Writers are economically marginalised. In a salaried working environment, an author would earn a minimum of around 60,000 euros in twelve months for an average of 1,000 to 1,700 working hours. This upfront investment by the writer is neither acknowledged, nor, when it comes to the use of European works by AI companies, even counted as a loss. But the fact is that AI developers are not only plundering the work itself, but also the economic and cultural effort that went into creating it — by the writer, the investments made by publishers, but also by the States when they launch programmes to foster literature, scholarships or residencies, tax breaks for writers, or prizes and other cultural promotion measures. AI developers invest in none of this yet profit from the working process and the final work. This free-riding affects not only the book sector but all cultural sectors whose investments are being taken away for commercial interests without their approval, financial redress and any transparency whatsoever.

The EWC attempts to achieve a mutual understanding among EU policy makers including the European Commission that various AI products are based on the non-consensual exploitation of our works, rendering our labour, were fruitless.

This, for the EWC, is a deeply troubling indicator: for this is no longer merely a matter of legal, economic or geopolitical disruption. The exploitation of writers in this way affects humanity’s sense of identity in an era of technological promises, an era of economic concentration amongst a handful of non-European players, and if affects the future of identities, language and shared values. 

In the context of the mass reproduction of products that simulate culture, the issue is how AI which is designed outside Europe is distorting Europe’s cultural fabric. Diluting it. And with it, everything we associate with culture in the sense of the arts, media, channels of information, and the formation of opinion and education. Culture is what surrounds us every day: it is socialisation and the daily glue that holds society together. It is the human element of being human.

Accordingly, Europe needs a clear, fully informed strategic response to shape its sovereign, resilient and distinct cultural and information space, and thereby significantly safeguard its very own fundamentals.

This requires political action on your part to ensure that Europe does not remain a passive ‘organ donor’ of cultural material for machines, but rather shapes the conditions under which such material may be used – starting with a stronger protection for authors. This, in turn, encompasses not only enforceable intellectual property rights, but the protection of working and living conditions, the structuring of contracts and negotiations, and their integration into equitable social systems.

Culture is everything that surrounds us. Accordingly, the CULT Committee has been entrusted with the most important task of recent years.

Six key issues to be considered within the Report by the Rapporteurs and the CULT Committee

2 Sovereign wording, confident stance: to know what AI is and what it is not is the key to regulating it

Gaps in the European policy Draft Ini Report  should be filled with a technically informed approach and vocabulary.

Most policymakers were quick to uncritically adopt the language used by the salespeople of non-European AI companies. A statistical technology that is, in essence, rather simple is exaggerated through the use of humanising terms like “learning”, “smart” or “understanding”. It adds to the hype around AI, whilst at the same time obscuring what is actually at its core:

“Artificial Intelligence” is a term without concrete object. AI neither refers to a coherent technology, nor a certain power, e.g. “intelligence”. Instead, the 70-year-old term functions today as an ideology that falsely attributes agency, intent, and authority to data-based statistical probability software.

 “AI” as a term and ideology was invented 1956 in a tender application to circumvent terms that felt pale – “Information Processing” and “Cybernetics” – and was used to examine how computational information techniques could simulate minor parts of human-like decision-making, behaviour, and action. The false premise which forms the basis of today’s AI-hype is to consider human brains and human behaviour as sort of linear information processing. There is no academic field that supports this theory.

The research on cybernetics and “human-simulating” information processing was strongly fostered by the budget of the US military over four decades. Voice assistants, face and environment recognition, autonomous decision making: these methods had been developed for war and military operations. From 2010 on, AI was given pro-tech and public interest narratives, which led to today’s reception of AI as an anthropomorphised IT, while it is, in its core, data-based statistics operating in high scale probability calculation, through the use of billions of copyright-protected works.

In our field, automated text software, like ChatGPT or Claude, reproduces linguistic forms based on statistical probabilities without any actual understanding of meaning. By contrast, human natural language is symbolic – pairing form and meaning. Numerous studies have demonstrated just how closely the output nowadays resembles the input: AI products are copies, plagiarisms, remixes, samplings. Our book works are reproduced by regurgitating parts of them, their striking ideas, their characters and their style. At the same time, the error rate is enormous, and since the “stochastic parrot”[1] can neither read nor check its own output, it also makes up events, court rulings, people and biographies.

Everyone must be aware that generative technology in particular – be it in the form of text, audio or audiovisual output – uses data-driven statistics to calculate the most likely outcomes, yet at the same time fails to recognise the meaning of a word. There is no core that can “think”, “learn”, “create”. Consequently, “humanised” vocabularies are whitewashing and obscuring the view of legal, moral, social and cultural impacts.

On the other hand, the label AI is often applied to technologies that have nothing to do with it: be it algorithms used for book recommendations or posts on an Instagram feed, activities which do not rely on probability calculations but on fixed rules, or software such as spell-checkers, Excel formulas, Photoshop or email filters. This is not AI, it is logic-based, static IT, which has one main advantage: it has not stolen from millions of professional and private originators to fill its data banks.

With this in mind, the EWC kindly ask for clear distinctions in what kind of technology is meant, and that discussions on AI try to avoid the marketing speech of AI companies. “Generative AI” for instance is works- and data-based pattern matching or synthetic text extruding software or generative automatisms; “training” is actually development or programming, “content” is generated output, while “works” shall be only human made.

We hope that the authors of the report will find their own language to employ confident, non-hyped expressions as an essential part of the desired European sovereignty while talking about “AI”.

3. The opportunities and challenges arising from the use of AI in the cultural and creative sectors:

Machine-generated products have no economic benefit for writers and translators – only for fraudulent actors.

The EWC analysis of 2023[2] already showed how the promised opportunity of AI is built on exploitation of labour hours, investment, human innovation, and how the output and books-imitating products harm remuneration, reputation, and reinforces bias, language discrimination, and leads to heavy de-skilling in reading, writing, and understanding. Over the past five years, translators in the five larger book markets saw a decline of commissions by a third, illustrators by a fourth. Authors are competing with AI-outputs on their market – as it is thought that 50% of new e-book titles are nowadays generated by machines – and have a loss of licensing revenues in the billions, while AI-products simulating books misuse privileges meant for human creation: revenues through public library money or device levies, prizes, scholarships, and other benefits meant for cultural assets created by human originality e.g. reduced VAT or fixed book prices.

If we assume that a writer spends around 1,000 hours per book, this already represents an average labour value of 60,000 Euro. As text generators attempt to simulate human language, they draw on both the authors’ working time and the finished book. Extrapolated across the millions of titles utilised from Europe, financial value amounting to hundreds of billions has been exported to the USA and China – and the resulting software is producing competing products.

Some concrete examples of growing challenges for writers in the book sector:

(a) Uncontrolled AI output is being pushed into the bestseller lists with click farms. For years, the global provider Amazon has been flooded with thousands bogus books by fake authors whose text products have been produced by generative software. Bots from organised troll farms “click” these book-imitating products, articulate random machine-made reviews, and push them into the bestseller lists.[3] This led to a rapid decline in revenue for human authors by shared-revenue models, such as Kindle KDP (a pool of revenue divided by pages read and number of authors, similar to Spotify). At peak times, 80 out of 100 Kindle KDP bestsellers are AI editions.

Amazon refuses to label AI output.

(b) Identity theft and name deception: The world’s most important review platform Goodreads, like Amazon, is flooded with automatically produced products published under the illegitimately used names of real human authors (or slightly altered spellings of known writers). These fake books are listed as new releases in the authors’ profiles and entice readers to buy them. The income from these bogus books flows to fraudulent actors. Human authors who are cheated out of their earnings must spend money to defend themselves with lawyers.

So far, neither Goodreads nor Amazon have stopped this identity theft.

Nearly every second new e-book is presumed to be produced by a chatbot, tricking buyers to believe it’s a human made work, and letting them pay the same prices as for real labour.

 (c) Unauthorised machine translations channel sales to fraudulent actors: We observed cases of books being illegally translated from, for example, English into Portuguese by means of robot translation without a licence, and published under a different name, usually in Amazon self-publishing. The revenues flow to unknown beneficiaries.

 (d) Illegal remuneration claims to collective management organisations (CMOs) and state funding programmes:Automatically generated and machine translated press articles and machine translated books, or even regenerative produced images, cash in on private copying levies from CMOs. Likewise, scholarships, prizes or state grants are – unintentionally – driven towards machine products instead incentivising human works. Machine “translated” books, and their non-authors receive library money or apply for prizes and state subsidies.

 (e) Amazon is producing non-licensed derivates of books and chatbot features.

The “Recaps” (“Story So Far”) feature summarises prior books in a series. How Amazon approached acquiring rights to copy and ingest full books to create generated summaries, has until now, not been answered. Kindle Translate is an “AI”-based translation, launched o6/11/25, offering machine translation from English and Spanish, and from German to English. This will impact translators’ job massively, apart from impoverishing, of course, the final outcome of the “translation” that the readers are going to have access to. The feature “Ask this Book” is, since 11^th December 2025, on the Kindle iOS app in the U.S. and lets readers ask questions about the e-books they are reading. Requests for information about licensing agreements for this feature and how to prevent mistakes, or to protect the text from further “AI” development, are still not answered by the tech giant.

Furthermore: self-publishers publishing their books at Amazon do not get any offer to apply a machine readable TDM opt-out. This concerns over 500,000 new self-published titles in the European markets.

(f) The tech companies Anthropic and Amazon are systematically buying European out-of-commerce works in bulk, shipping them across the Atlantic, digitising them and destroying the books afterwards. Apply a TDM opt-out? Never possible. Transparency? Please. Is this against Art.8 CDSM Directive? Indeed. And now?

In vast storage facilities in the U.S.A., machines are cutting off book spines as if they were slices of Chouriço salami.

Scanners digitise pages in bulk. Within a second, works on which European writers had spent years writing end up in the digitised cloud: as source material for generative text software.

The scanned pages are shredded and turned into toilet paper.

Under the chiffre “Project Panama” Anthropic was pursuing a secret plan: to “destroy” books through scanning worldwide. This is evident from an internal document that was made public at the end of July 2026 as part of legal proceedings. The same is done by Amazon, as an investigative report in August 2026 revealed, in Las Vegas, by a company with the logo of a Tyrannosaurus Rex slaughtering a book. Amazon is alleged to do the same to rare books and out-of-commerce works from Europe, using them for their so-called AI systems. The news outlet 404 Media found evidence supporting a theory that AI firms are systematically working through lists of ISBNs to make sure every unique book gets scanned and is destroyed.[4]

Conclusion: Generative software benefits fraudulent actors and non-European companies the most.

People who have never written before are using automated text software to churn out books and are not required to label it human readable – whilst authors are forced to use AI against their will to keep fees down or are wrongly suspected of having used AI, and as a result lose prizes, reputation and contracts. How will we prove in future what is genuinely human? If machine-generated content is not consistently labelled as such and excluded from remuneration streams, those copy-machines will enjoy more advantages than humans, who will find themselves having to helplessly defend the norm – human-created work.

 

4. The protection of copyright and related rights, including transparency regarding the use of protected works for AI training and the fair remuneration of creators and other rightsholders:

There is currently no regulation on the side of authors, artists, and the cultural and creative sectors. International treaties on how exceptions shall be implemented are ignored, including the 3-step test.

Several studies[5] and legal analysis[6] have proven that the Text and Data Mining (TDM) exception(s) Arts 3 and 4 of the CDSM Directive (EU) 2019/790 were not intended by the legislators for the development of (generative) AI. TDM is neither on a technical basis nor within legal implications equivalent to any AI-related processing and does not cover the developing of profitable technologies based on data-driven statistics.

This excessive overinterpretation has led to the loss of billions in investment by the European cultural sectors, exploited by non-European companies to build software whose outputs are competing with those works the companies have availed themselves to. What has remained forgotten in EU policy, however, is the fact that the Union’s framework must follow binding international treaties, particularly the three-step test, which, as Nicola Lucchi[7] explains, is “requiring that exceptions do not unfairly conflict with copyright holders’ economic rights. Any proposed reforms to copyright policy must therefore fit within these constraints, balancing innovation with rights protection.”

The current EU policy framework, although with partly good intentions, does not offer solutions, and even actively hinders the enforcement of authors’ rights.

• Although the AI Act demonstrates a willingness to address transparency, its provisions regarding transparency of the input – material digested by developers – and labelling requirements are useless when it comes to checking and enforcing our rights. “Watermarking” is not sufficient when it comes to aspects of being fair and transparent to consumers, to collective management organisations, or to authors, in examining how similar any output is to their original works. Plus, start-ups are spreading with the promise that they can remove any marking within AI products.

• Neither the input template by the AI Office is sufficient, nor the labelling regimes, with too many arbitrary exemptions meaning rightsholders cannot address the above-mentioned challenges. The entire benefit lies with AI developers, whilst we are expected to bear the burden of proof without access to the evidence.

• Apart from the fact that the TDM exception is being misused to cover the development of commercial generative software for five years, no comprehensive opt-out systems have been put in place, nor, where they do exist, are they being respected by AI developers.
• Meanwhile, the markets are flooded with AI slop products (see the case studies above). The trust of readers and cultural consumers is waning. Authors are forced to justify why they do not use AI – or to defend themselves against accusations that they have used it; this is increasingly affecting people of colour and ‘own voice’ authors.

Therefore, the EWC calls upon the Rapporteurs to promote authors’ rights and copyright:

• By seeking remedies for the damage caused by copyright infringement and holding AI providers, deployers, collectors, accountable and liable. This would also require the preparation of a study on the damage caused to the authors as primary rightsholders. Such study should not be financed by the affected stakeholders. but rather carried out at EU level. The advocacy of the Members of the European Parliament would be most welcome here.
• By calling upon the European Commission to remove the non-applicable TDM exception(s) within their reform of copyright, to allow entirely voluntary and individually negotiable licensing by authors against appropriate and proportionate payments, without posing any further “AI exception” or mandatory licensing upon authors and further rightsholders. We kindly ask you, although this might not be in the very upfront territory of the CULT, also to remind the European Commission about the lack of the 3-step-test: generative software is heavily conflicting with our legitimate economic rights.

IV.2       Safeguard trust by full transparency as key principles – for readers and for authors

Generative technologies and their current usage undermine public trust in democratic societies, including trust in news, media, and the provenance of creative and cultural works and books. When automated generated output circulates without human readable disclosure and labelling before a purchase process is made, it becomes impossible for readers to distinguish between human-created works and machine-generated products. Thus, they cannot make informed decisions before purchase, or are being manipulated into accepting false premises, or suffer reading uncorrected pseudo-facts. This could potentially weaken confidence and trust in cultural creation and distribution overall.

The people of Europe deserve to know what they are reading, listening to, believing, or purchasing. Also, non-European online platforms need to be obliged to label machine outputs as such with a fine-granular approach, e.g. like an energy diagnostic score, and what kind of technology was used (or not). The most immediate labelling should be of generative and manipulative techniques, but in the long run also where decisional algorithms have decided what the user should see, read, or buy.

The CULT committee ought to have in mind the European Parliament resolution of 14^th September 2023 on the Future of the European Book Sector (2023/2053(INI)) (Rapporteur Tomasz Frankowski), emphasising the essential role of books in our societies, and “the importance of transparency related to AI training, including data collections and their sources”.Hence, transparency obligations – such as the envisaged in the context of the Digital Fairness Act, mostly aimed to prevent unfair or manipulative digital practices, which would also be the case in the cultural and media sphere – are not only technical requirements, but essential instruments for protecting democratic values, cultural integrity and fair cultural market conditions in the age of disruptive technologies. As the EWC has been claiming in the consultation of the Code of Practice on marking and labelling of AI-generated content[8], full transparency regarding AI-generated products is essential.

We call upon the CULT to draw up an ambitious report, and to put transparency at its centre:

• Transparency of the Input. Usage documentation by any works-using developers, commercial and non-commercial, shall include sources, acquisition, exploitation of protected book works and title by title. The current “template” born from the AI Act, is a perfect instrument to avoid any senseful transparency, and hinders writers and translators from actively examining and exercising their rights.
• Transparency of Output. Also, all machine created products imitating books shall be labelled without exemptions as currently presented in the Code of Practice on Transparency of AI-generated Content (AI Act: Art. 50) by the European Commission, e.g. for generated “fictional” or “poetic” output, as these are exactly the competing products betraying consumers and harming the markets and human authors.
• Labelling should be mandatory and continuous across the entire cultural value chain, operating at every stage: within a data-driven model; during inference and re-prompting; file conversions; and downstream distribution. This persistent marking chain must start with the model and continue through the user to the recipient, to build an unbroken provenance history and prevent the removal or the degradation of labels.
• In order to achieve that persistence, labelling should be human readable and fine-granulated. We advocate for a multi-approach system, which combines marking, detection and provenance certificates, since there is no single technique that works for all formats. Here, we are concerned about the primacy of watermarking as the main option of this Code of Practice on Transparency of AI-generated Content, since it is easily erasable within texts, and currently does not come with any detection software to identify red flag markings.
• Additional reliable Provenance labelling will be especially needed to verify press and media works, to enhance trust in media, and the fourth estate of democracies.
• Another combination we stand for is that both human-readable labels and machine-readable marking are necessary, in the sense that marking and labelling should be useful for diverse profiles of stakeholders and the different degrees of transparency they need. Authors and rightsholders need to prove the human origin of their works for licensing and remuneration; platforms offering books need metadata for correct payments and categorisations; readers need trustworthy information to decide which cultural goods they want; CMOs require such labels to license and remunerate properly the authors; and states will want to avoid granting awards, cultural subsidies and tax privileges to AI-generated products.
• It is imperative to introduce obligations on digital platforms to accurately label AI-generated outputs as such, beyond what is established by the AI Act’s obligations, which only apply to AI system providers and deployers. This could also be promoted by guidelines.

Regulations on compliance and transparency in public broadcasting, tele-media providers and the cultural sector regarding provenance, responsibility, authorship and respect for the rights of third parties have been implemented and applied throughout the cultural and media sector for decades. There is no reason why the same rules should not apply to the use and distribution of machine-driven applications, especially when they simulate literature, culture, news or media.

5. The impact of AI on creative professions, working conditions, skills and employment:

V.1.  Europe needs a real assessment of the market harm of generative and manipulative AI on human works, the primary markets, their authors, deskilling and intergenerational inequity. The EWC calls for the Ini-report to initiate a comprehensive and EU-wide study on market harm done under the headings of:

• Input: licensing revenue and investment loss through non-authorised use of millions of works, related to writers – the book sector – and state subsidies;
• Output: income declining, both through job loss, and by the AI-driven marginalisation of professional writers’ and translators’ skills, as they are more and more forced into using generative or manipulative AI, e.g. post-editing, reproducing visual AI outputs, or following a ChatGPT-made “exposé” for a novel or other text, and being demoted to mere “maintenance staff” for technologies of poor quality, but especially for significantly reduced fees; subsequent loss due to emerging “AI features” within cultural works, using books and other art without licensing, in the case of derivates, again competing with human authors;
• Overall: impact on collateral consequences, including on states and European societies, e.g. reduced tax payments, a decline of payments into pension schemes or social insurance systems and further health care (stress related, especially), societal, and culture-economic aspects.

 A comprehensive study could give an overview on the extent of the harm done and therefore serve as a basis for the common reflection on future EU legislative intervention. It would close an important research gap – because most decisionmakers only see unproven promises made by so-called “AI” and ignore the actual, devastating effects on authors as the source of Europe’s third-strongest economic sector, and the resulting state budgets. Likewise, an in-depth examination helps to understand the core aim of the 3-step-test, and to not grant exceptions to the detriment of authors.

Consequently, we ask the Members of the European Parliament to advocate for the necessity of such a comprehensive study towards the European Commission and the Member States.

V.2  Develop checks and balances on the professional and social situation of authors regarding the pressure coming from automated machine products cannibalising the human market – including through funded projects by the EU, and misuse of cultural asset-privileges by machine products

AI and its destructive impact on culture and its sources, the authors, should be an important aspect within the follow-up of the High-Level-Round Table and the Commission’s plan on this issue, including regarding the planned “European LLM”, and the “AI Apply Strategy”.

Public funding, also within European programmes, likewise need human-centred governance regarding automated text and other cultural works simulation. Human-made works of writers and translators should still be priority in public funding safeguarding diversity, social responsibility, and human creativity as keystones of society. The current over-promotion of AI throughout EU funding is harming the sector and authors even more. For example, the TV-channel ARTE is encouraging the development of AI within the EU Commission funding programme, therefore endangering at its core human works and creativity.

Indeed, Member States enjoy a certain national flexibility in fostering and promoting their creative sectors and promoting cultural assets; for example, by reduced VAT, by the fixed book price, by a basic income for artists (e.g. in Ireland), or by social insurance systems where the state pays an additional contribution to the payments to authors. Calls for tenders or commissions, or the funding of festivals, prizes or similar subsidies, should be subject to clear conditions to promote human made work, including the obligation towards large media online retailers or content providers, to label AI output and to avoid consumer betrayal.

Therefore, the EWC urges the CULT do draw attention to these aspects within the draft report:

• As machine products, especially when unlabelled as such, commit fraud and misuse privileges meant to foster human works, the EWC calls upon the Committee and Members of Parliament, to encourage Member States to set up binding rules against fraudulent AI products.
• We urge the CULT Committee to call on the European Commission to ensure that human creativity is put front and centre in the upcoming AI strategy for the CCS, as well as any upcoming action under the Culture Compass. This should entail including CCS stakeholders being active participants in AI governance, especially in relation to policies directly affecting them.
• We call upon Members of Parliament to advocate for the establishment of rules to avoid the funding of non-human works made by AI-based products, as we see a need for public funding to be channelled exclusively towards human-made works in the book, text, and cultural sector.

V.3.   Stop the rapid rise of de-skilling in culture, education, and for the next generation of authors

According to Open AI’s own study,[9] 800 million people, predominantly young people between the age of 18 and 25, submit 18 billion queries worldwide every week to the error-prone, pre-censored and resource-consuming communication simulators. People no longer write or read themselves either but have teaching materials and books summarised by probability software. University lecturers in Europe and across the Atlantic[10] are increasingly encountering the phenomenon of young people who are not used to reading and who find it mentally, emotionally and physically exhausting to read texts longer than a chat. The cultural techniques of “reading” and “writing” are not particularly high on the agenda in European national curricula and Europe still suffers under the consequences of a decades-long education crisis. As is so often the case, AI only acts as a catalyst for emerging disasters and is a magnifying mirror of existing desolate conditions. Including rapid de-skilling: The substitutive effect of generative and manipulative software is leading to a loss of skills that not only threatens writers, but society at large, with consequences for the youngest generation, for education and for all values we see in cultural diversity – but also onto the freedom to form one’s own opinion.

As European competitiveness is more often on the agenda than solutions to the educational crisis, we emphasise that the importance of reading skills is also highlighted in the Draghi Report as well. Indeed, a deeper understanding of texts is vital for lifelong learning or quick adaption to the various changes in the job market. To protect the future of these skills, and to avoid intergenerational inequity, the EWC endorses a 2023 call of the EP on the Commission and Member States to develop:

“Job creation action plans and sector-specific financial support and measures to ensure access to adequate social protection for those affected by digitalisation and AI-related job losses”.

The EWC asks that the CULT Committee reiterates these calls within its own-initiative report and demands the Commission to put forward decisive measures to remedy the impact of disruptive technologies on cultural metiers and skills. We propose, to emphasise the need for advanced media literacy, and not only to ask: is this true? – but to ask which system produced this and why.

V.4   Freedom of art begins with the freedom to choose one’s work equipment

The working environment of writers and translators is becoming increasingly complex and even more unstable. With the EWC AI Toolkit for the Book sector [11] (2024), we have drawn up necessary, sector-specific codes of conduct to be adopted via a mutual understanding of a shared responsibility among authors, publishers, retailers and other institutions and individuals, such as libraries, reviewers, and also agents and the reader themselves. This includes transparent communication and respecting the moral rights of authors to the integrity of their work. Contractual partners, for instance, should always seek an author’s approval before using generative AI in relation to their works (for cover art, audio books, or for translations), and respect their rejection. The same goes for libraries, to not feed digitised works into unsecured GPTs to get a summary, unintentionally feeding an LLM overseas.

On the other hand, authors should be protected against false claims that they have used AI in their writing.

Our goal is to establish a mutual understanding on the utilisation of different kinds of technologies and software, such as, but not limited to, synthetic voices in audio books, machine translation, generative cover illustration, and any other adaptation of the work by generative works and data-driven statistics. We are receiving reports of instances where the author is forced to accept AI-adaptations. For example, publishers are increasingly using AI-generated covers and when the authors request human visual artists, the authors are required to pay for this. If they can’t, readers react with the accusation, that, if a book has a machine-made cover, the author is probably also not human. Authors, as traditionally weaker contractual partners due to the lack of enforcement instruments by law, hardly have the alternatives to refuse this and are forced to live with a declining reputation beyond their control. Authors are pressured into accepting a technology that harms their own profession at its core – also by political means.

We consider it fundamentally inappropriate that, through political and costly EU initiatives such as the Apply AI Strategy or the targeted funding of machine-generated output, a technology that is morally, ethically and socially questionable is being forced upon us.

Authors should have the right to choose to use human work in relation to their book, and refuse AI generated adaptations, without being disadvantaged and without negative consequences, such as lower royalties. Therefore, the EWC hopes that this report draws attention to better contractual practices:

• We encourage the report drafters to call upon Member States to set up mediation entities for authors who wish to say “no” to the use of their works within TDM, GPAI, or generative AI usage, or who have problems with their publishing partner, when the author rejects any adaption of their work by machine generated software, e.g. synthetic voice audio books, machine translations, or generated covers.
• Particularly in the rapidly changing AI-ideology era, in which our works are being used or adapted en masse and without consent both within and outside the sector, it is of utmost importance for Members of Parliament to initiate an adjustment of EU competition law beyond the scope of the report. Authors’ associations must be finally allowed to publicly advise and provide information on contracts, clauses, remuneration, and how to deal with unauthorised use and adaptation.

6. The preservation of cultural and linguistic diversity and the visibility of European works in the digital environment.

As recognised by texts such as the European Charter for Regional or Minority Languages and the Universal Declaration for Linguistic Rights, Europe’s linguistic diversity is not only a core part of its cultural heritage, but also an essential pillar of democracy, pluralism and social cohesion.

However, the fast expansion of generative linguistic simulators poses significant risks to this European diversity:

1. a) Generative linguistic systems, and especially so-called LLMs, are built on material belonging to the dominant world languages, making these far more effective and with a stronger digital presence than less-used languages. This creates disadvantages for speakers of smaller or minority European languages. Over time, this imbalance could encourage a shift toward dominant and non-European languages in publishing, education, business and online communication, threatening both cultural diversity and the linguistic rights of minoritised communities in Europe.
2. b) Generative text automations accelerate cultural homogenisation in general. In the creative and literary fields, linguistic pattern matching software optimises for mainstream, market-tested narrative patterns, reinforcing dominant cultural models and marginalising local and unconventional voices. If widely adopted in text generation, “translation” or “editing”, this data-driven probability software could standardise language use and pressure the unconventional authors to conform to dominant norms set by machines, undermining not only linguistic diversity but also the cultural expressions embedded within it. Generative AI, and its pre-censored vocabulary, coded within non-European oligopolies, do not equitably represent local expressions, terms, idioms and variety of opinions – especially in multilingual, culturally diverse markets like the Union.
3. c) The massive use of machines for “translating” texts implies damages to cultural diversity and a remarkable linguistic impoverishment, since machine “translation” often prioritises semantic equivalence over cultural nuance. Idioms, oral traditions and culturally specific references risk being simplified or neutralised, potentially leading to a gradual erosion of Europe’s rich plurality of voices, principles, identities and visions. Plus, as translators are replaced, the de-skilling and involution of cultural techniques multiply.
4. d) Static algorithms by global monopolies like Amazon decide what cultural works are promoted and purchased. Today, cultural works find consumers, rather than the other way around: Algorithm-driven feeds, recommendation engines, and user data-powered personalisation have fundamentally disrupted how people engage with culture and make purchasing decisions.

As a result, decision-making code by non-European entities define European spaces with an “Economy of Attention”, in which certain kinds of cultural works are over-presented, while others sink into what we term a “Dante’s loop of invisibility”.

Readers or consumers of books, of music, or film cease to actively search to discover art, and instead now passively view a customised proposal created by machine – to the detriment of niche topics and avant-garde, subversive or non-mainstream cultural works, and to the disadvantage of local retailers and the whole European sector.

 

To remedy these impacts, the EWC kindly requests to the Members of this Committee, to continue pursuing the intentions expressed within the European Parliament resolution of 14^th September 2023 and on the Future of the European book sector (2023/2053(INI)), describing the role of books in our societies as “an invaluable source of knowledge, education, culture, information and entertainment and a vital means of preserving and disseminating the EU’s values, cultural and linguistic diversity and cultural heritage.” 

The further this aim, the EWC proposes the following measures to prevent the homogenisation of European linguistic and cultural diversity in the literary fields, and to re-establish discoverability:

• Stimulating the creation or consolidation of funding programmes for creative and for professional writing in Europe, especially for works written in medium-sized and small or lesser-published languages;
• Stimulating the creation or consolidation of funding programmes for literary translation among Europe’s diverse languages, particularly medium-sized and small or lesser-spoken ones, to maintain the current quality standardsand to foster dialogue between different literary and linguistic systems on equal terms and with all the necessary linguistic and cultural nuances.
• Promoting exchange and intercultural dialogue programmes among the different literary systems of the European Union, involving not only authors and translators but also publishers and booksellers as agents in the book value chain, but foremost the federations and associations as key hubs for capacity building.
• A legislative work on active anti-discrimination of the discoverability of European cultural works, like the AVMS Directive. With the Audiovisual Media Services (AVMS) Directive (amended in 2018, implemented in 2022) new obligations to promote European cultural works for on-demand services, the European decision-makers made a decisive move to foster European values of unity in diversity. For example, on-demand media service providers in a national jurisdiction must offer at least a 30% share of European works in their catalogues and ensure the prominence of those works. The EWC calls on the CULT committee to propose in its Own -Initiative Report a similar legislative instrument, to foster the cultural ambitions of EU Member States, to protect diversity and to enhance the discoverability of European cultural works in all fields, where AI-driven agents, systems and static algorithms pre-dominantly override linguistic and cultural plurality and suppresses European works and languages.

 

VII. Possible policy and regulatory measures to support responsible innovation and the sustainable development of Europe’s cultural and creative sectors.

We can understand the approach of pursuing a competitive ‘in-house AI’ strategy to achieve European sovereignty and independence from dominant U.S. or Chinese tech companies, which treat Europe as a self-service buffet of data.

However, we cannot fully endorse such a strategy.

Europe’s competitiveness does not lie primarily in simply replicating the same model, albeit with a few more legal safeguards. Europe’s competitiveness lies in the future of what is – for the time being – its third-strongest economic driver: the creative and cultural economies.

Accordingly, our approach is not to assume that a ‘balance’ between technology and culture can be achieved. That balance was lost long ago; we have been the ones losing out for three decades; to restore it, regulations would have to be introduced exclusively in favour of culture, without regard for commercial interests.

The current so-called ‘AI’ development is not an innovation either; it is a 70-year-old theory in new clothing and with more IT power. It is only ever as good as the works and data on which it relies.

We call for a change in Europe’s mindset. We have everything – and much more besides – that so-called generative ‘AI’ cannot achieve in this century: courage and determination, cultural distinctiveness and new ideas, and a political will to shape our future together, which advocates for a sense of ‘us’. Accordingly, political regulations should focus less on how to relate to AI. Regulations should impose more regulation of AI technologies, and offer culture more – enforceable – rights, as well as foster the living and working environments of the individual originator. This can be enabled by:

The most important priorities for policy and regulatory measures:

• Call for an EU wide study on the market harm and socio-economic impact of so-called generative AI;
• Seek remedies for the damage caused by copyright infringement;
• Remove the non-applicable TDM exception(s), to allow voluntary and individually negotiable licensing by authors against appropriate and proportionate payments, without posing any further “AI exception” or mandatory licensing upon authors and further rightsholders;
• Encourage Member States to set up mediation entities for authors who wish to say “no” to the use of their works within TDM, GPAI, or generative AI usage;
• Initiate an adjustment of EU competition law, so that authors’ associations are allowed to publicly advise on contracts, clauses, remuneration, and how to deal with unauthorised AI adaptation;
• Set up a human-readable labelling system for AI-fabricated products of all kinds;
• Request the initiation of a legislation on active anti-discrimination for the discoverability of European cultural works;
• A dedicated Directive for the social protection of authors, artists, and cultural professions.

 

Final remarks

In a rapidly changing world where European values, such as diversity of culture, are under pressure, for example by non-European AI-Tech or autocracy-driven book bans, carried out under the usage of AI recommendations, the European Union needs a distinctive, dedicated, and committed own-initiative report to strengthen European cultural sources, the authors, as the most needed pillars upholding European values and cultural diversity. Authors are the critical infrastructure of the European society.

The writers of Europe set their hopes on the CULT Committee to draft an ambitious report that stands firmly with Europe’s writers’ community in addressing the challenges imposed by advanced technologies based on our works and labour.

With this in mind, we kindly ask you to take our submission into consideration. We remain available and are eager to support the Rapporteur and Co-Rapporteurs in their tasks to the best of our ability.

 

With our kindest regards:

Nicole Pfister Fetz                    Sebastià Portell Clar               Nina George                                      

Secretary General                      President                                  Commissioner for Political Affairs      

Cited sources

[1] Emily M. Bender, Timnit Gebru, Angelina McMillan-Major and Shmargaret Shmitchell: On the Dangers of Stochastic Parrots:Can Language Models Be Too Big?, in: FAccT ’21: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 2021, p. 610-623,

[2] https://europeanwriterscouncil.eu/gai-is-based-on-theft/

[3] https://www.vice.com/en/article/v7b774/ai-generated-books-of-nonsense-are-all-over-amazons-bestseller-lists

[4] https://www.404media.co/we-tracked-a-shipment-of-rare-books-it-ended-at-an-amazon-ai-training-facility/

[5] Tim Dornis, Sebastian Stober; https://arxiv.org/abs/2502.15858, 21 Feb 2025 (v1), last revised 17 Mar 2026 (this version, v2)], Generative AI Training and Copyright Law

[6] Generative AI and Copyright. Training, Creation, Regulation. Nicola Lucchi, PhD – Serra Hunter Professor of Comparative Law, University Pompeu Fabra, commissioned by by the European Parliament’s Policy Department for Justice, Civil Liberties and Institutional Affairs at the request of the Committee on Legal Affairs, https://www.europarl.europa.eu/RegData/etudes/STUD/2025/774095/IUST_STU(2025)774095_EN.pdf

[7] The three-step test in international copyright—a global framework for generative AI training, by Nicola Lucchi, Tim W. Dornis, Pascal T. Sierek. https://onlinelibrary.wiley.com/doi/10.1111/ablj.70022, by 5 August 2026.

[8] https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content#1720699867912-0

[9] https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f142e/economic-research-chatgpt-usage-paper.pdf

[10] https://www.theatlantic.com/magazine/archive/2024/11/the-elite-college-students-who-cant-read-books/679945/

[11] https://europeanwriterscouncil.eu/ai-tool-kit2024/