Artificial intelligence, sometimes perceived as an opportunity, at other times as a cause for concern, is kindling numerous debates in the publishing sector. But what kinds of upheavals is it bringing about exactly? The professionals of the francophone Belgian book industry, far from adopting a wait and see policy, are readying themselves to anticipate the upcoming transformations. Decryption.

In the Wallonia-Brussels Federation, the publishing sector amounts to some 10,000 indirect jobs in Brussels and Wallonia, according to various studies. ‘The main links in this chain are the authors, the publishers, the distribution and the bookshop owners. At each stage, there are dozens of professions which may be impacted by AI. Of course, you also have to include the readers, who are at the end of this chain,’ explains Geoffroy Wolters, delegate general for PILEn, a community platform which offers aid to Walloon and Brussels book industry professionals faced with technological mutations.

A heterogenous perception amongst the publishers

In 2023, francophone Belgian publishers generated a turnover of 327 million Euros, according to the Belgian Publishers Association (ADEB). The principal publishing productions were pre-empted by the comic book, the human sciences, school books and children’s literature. The perception of AI amongst these publishers nonetheless remains heterogeneous. Carine Lecomte, deputy director-general of the ADEB (which numbers around 85 members for 130 publishing labels), is already observing polarisations. ‘We have just established an AI working group within the ADEB and I can already say that there are very varied positions. Some see AI as an opportunity and have already integrated it in their daily practices and in their business proposals. That is true of legal and school sector publishers who took the digital transition plunge some time ago,’ she explains. A large proportion of their turnover in fact stems from digital tools (applications, internet sites, databases, etc.). Generative AI – the LLM (Large Language Models such as ChatGPT, Gemini, LLaMA, Claud, etc.) or image generators (Midjourney, Stable Diffusion, DALL-E…) – will enable them to add value to their content and to create new reactions with the readers. It is with this in mind that the legal book publishing group Larcier-Intersentia, based in Brussels and Louvain-la-Neuve, recently launched GenIA-L, a tool which enables responses to be generated as regards legal matters by linking legislation, jurisprudence, doctrine and recommendations arising from various databases. The publishers of school books will also be able to ‘generate exercises adapted to learners or by illustrating certain pedagogical content,’ emphasises Carine Lecomte.

In contrast, other publishers are purposefully staying in the background of the generative AI phenomenon. This staying on the sidelines in particular affects original publishing productions. It is therefore not surprising that the giants of the graphic novel – Casterman, Dargaud-Lombard and Dupuis – as well as the publishers of children’s literature, have not yet really invested in generative AI. ‘For the moment, these publishers are preferring to observe the impact of AI and their primary concern, at this stage, is to maintain their authors’ copyrights. They will probably only take action in the coming months, depending on developments in tools, the sector’s practices, the legal framework,’ notes Carine Lecomte. Finally, generally speaking, small publishing houses feel little, or not at all, affected by the advent of generative AI. This is not preventing the emergence of structures such as Stripik, a Liège based publisher who has designed a free-of-charge application aiming to generate illustrations for comic books. To conclude, let us observe that, in parallel with generative AI, other technologies are being developed by the major publishing houses: ‘data publishing, the management of stocks, the detection of plagiarism, the evaluation of script (Editor’s note: see the ActuaBD article) or the optimisation of print runs are solutions which can make use of AI and which are likely to be of interest to the publishers,’ explains Carine Lecomte, before adding, ‘the problem is that, in francophone Belgium, 80% of the publishing houses are (very) small or medium enterprises and do not necessarily have the skills or the means to avail themselves of these technologies.’ A reality which may accentuate the variances between publishers on a market where the most powerful have in particular the power to standardise the sales price of books.

Tensions ahoy for the authors

Turning to the authors, for their part, the issues are just as vital. ‘The authors are in no way anti-technology. In general, they are even early adopters and are interested in generative AI. Then again, the real question is working out what the professional norms are and the plus value for them,’ sums up Frédéric Young, delegate general for Belgium at the SCAD (Society of Authors and Composers of Dramatic Works) and delegate general of SCAM Belgique (Society of the Authors of Literature and Documentaries), an organisation which brings together some 4,000 authors resident in francophone Belgium. The finding shared by the majority of book industry professionals is the following: whilst original creations risk being less affected by the arrival of AI, the most conventional forms should be strongly affected. ‘One can for example imagine in the domain of children’s literature that AI will be quite commonly used for early learning activity books, with the shapes and the colours.’ In the same manner, a standardised literature, such as airport novels, will probably be impacted. Nevertheless, Frédéric Young prefers to maintain an overall perspective of the cycle of creation: ‘it is important to understand the business model of artists. The creation of original works is not remunerative in relation to the time spent. Research, trialling creative works, etc. in the jargon this is known as invisibilised time. The authors can allow themselves this low renumeration because in parallel they have other contracts where they have to accomplish standardised tasks which are more remunerative. Yet these tasks will henceforth be handed over to AI. That disrupts the already precarious balance of creation.’ Without immediately taking a position on the business model of artists, Marnix Verduyn, better known by the name of the artist Nix, is embarking on a doctorate in the Faculty of Engineering Science at Leuven, where he is working on the intersection between AI and the comic book. He writes on his website: ‘There is really no valid reason to deprive people of their incredible creative work. As an artist myself, I can’t justify advocating for such a change. That’s not what my research is about. Instead, I’m focusing on the parts of the creative process that are monotonous and repetitive, which I believe could benefit from partial automation. In discussions with fellow artists, I noticed a consensus that AI doesn’t have to pose a threat as long as we keep humans at the forefront of the creative process.’

Additionally, several professions find themselves directly threatened by the emergence of generative AI. Top of the list, the translators. The considerable progress achieved by the LLMs in effect enables an increase in the number of translations at lower costs, jeopardising a profession already below the waterline of the viable minimum. An impact which will by contrast be beneficial to the publishers who will in this see an opportunity to increase their transfer of copyright at lower costs, which risks being a particularly sensitive subject in the author and publisher relationship. Another potential risk: the democratisation of generative tools could be synonymous with the transfer of production tasks to be performed by the authors. ‘There was a time when the page layout of a file was handled by the publishers. As of now, authors are increasingly being asked to a provide a file which is ready to print. This transfer to the authors of a whole series of technical responsibilities risks accelerating with these generative tools and raises questions about the sustainability of the economic model,’ analyses Frédéric Young. Will the authors have to themselves generate the covers or the illustrations for their books? That remains a speculative hypothesis which will need to be observed attentively. In any case, Frédéric Young prefers to see this context as an opportunity for the sector: ‘as AI can have major effects on the employment of thousands of people, we need to pair the debate on the issue of the social protection of authors and take advantage of this wave to establish a professional status equal to the issues.’ 

Uniting to defend author’s rights

In any event, the publishers and the authors are generally in agreement on one point: the need to protect the creations from the giants of generative AI which lean on data extracted from websites or social networks to train their AI models (a technique known as web scraping). Frédéric Young reports on the stance adopted by the authors: ‘a large majority of them is shocked by the savage appropriation of their works so that models can be trained. Their voices need to be heard and this predation must be stopped.’ A favourable context in the European Union with the adoption of the AI Act which will come into force in 2026 and the exercising of the right to opt out provided by the Copyright Directive. This right consists of the possibility of a publisher of content to forbid this content being accessed by algorithms which would ‘mine’ the data to train themselves. A solution which is simple on paper but more complex in reality. ‘Our legislation very narrowly permits the harvesting of data for commercial purposes, but does so very openly for scientific ends. What concerns us is data used for research and teaching purposes and the loopholes which it may contain. We want to remain extremely vigilant as regards the threat of the use of the legal deposit managed by the Royal Library (KBR) in order to supply AI in the guise of teaching or scientific research,’ worries Carine Lecomte.

The consequences of this model, in short, weigh as much on the publishers as on the authors. ‘The AI phenomenon is macro but there is a micro impact on all of the stakeholders of the book chain,’ shares Frédéric Young. ‘If the real cost was passed on and you included the work carried out, in other words, the research, the creation, the publishing, the translations of a work which will serve to train the AI model – the creation of synthetic content would not be as advantageous from an economic point of view,’ adds Carine Lecomte. An unfair competition which it will be necessary to take a stand against. Even if it means imitating the decisions taken by the Financial Times and Le Monde, which recently signed partnerships with the American giants, agreeing to, in return for a fee, the use of their data? In any case, in view of the influence of the behemoths of AI, the response could be no other than collective.

Distributors and readers scarcely affected

Paradoxically, whilst AI is already triggering upheavals amongst the publishers and the authors, the distribution considers itself still minimally affected. ‘The independent bookshops are still involved with the issue of e-commerce, they are at a considerable remove from the problems of AI. The same is true of the libraries, apart from perhaps large organisations such as the Royal Library, which already has an AI solution to save time at the pre-cataloguing stage,’ assures Geoffroy Wolters. 

A relative permeability which is also perceived amongst the readers. ‘The first AI works are only now coming onto the market, and we are attentively monitoring the reactions of the readers. But for the moment, I don’t believe there have been any real reactions,’ continues the delegate general for PILEn (Partenariat Interprofessionnel du Livre et de l’Édition numérique). Yet the experiencing of a book is tending to extend to digital options (read the article published on kingkong). ‘Eventually, one could very easily imagine content which can be personalised by the readers,’ notes Carine Lecomte. At the top of the list, the audiobook which could doubtless adapt the voice of the reader with a deepfake of the author themselves or a member of the family to read a children’s book. ‘The day we get to this type of augmented content, that will be when the perception of the readers changes,’ concludes Carine Lecomte. A double-edged perception in which the consumer will have to be in a position to make an informed purchase concerning works which have or have not been produced by AI. ‘Especially given that in this AI context, the writing of a book “in the style of” is relatively easy to carry out. It will be necessary to avoid any confusion for the readers. With this in mind, the publishers are considering the issue of a guarantee of authenticity or a labelling so that consumers can identify and value original works as opposed to artificial productions,’ shares Carine Lecomte.

A study to detect needs and plan for action

One case file amongst others which will require genuine collaboration between the actors of the book ecosystem. While waiting for positions to be adopted as regards concrete actions, the time is ripe for strategic reflection. It is within such a context that in September 2024 there was launched an ambitious study, intended to run for 12 months, led by the PILEn. ‘This study was commissioned by the Minister for Economic Affairs. Our aim is to understand the perceptions, the opportunities and the stumbling blocks of AI in each link of the francophone Belgium book chain. We will work with territorial partners such as KIKK in Namur, the ADEB and the SCAM. We will carry out a benchmarking of the solutions in neighbouring countries, Anglo-Saxon and francophone, and not solely on generative AI,’ explains Geoffroy Wolters. The actors of the book sector will be able to, on the basis of this study, gain greater awareness of the opportunities AI may offer. This study will subsequently lead to the establishing of recommendations for political representatives to ‘implement a dynamic and coherent policy.’

Whilst it is still too early to envisage training modules dedicated to AI, the issue will very quickly turn up on the agenda of PILEn and for all the representative bodies of the book sector’s professionals. ‘At the ADEB, we have established our first priority to be the legal aspects related to AI: contractual clauses and good publishing practices will have to be recommended to our members. First of all, how to improve market expertise, the best possible knowledge of the needs and behaviours of the readers. Then the technical aspect is essential. For these two aspects, we will have to rely on more competent bodies, such as Digital Wallonia, for example, to provide our members with support,’ shares Carine Lecomte. As regards the authors, the SCAM has already begun to include AI in training modules. ‘For us it is urgent to train our members in generative tools. The SACD and SCAM Belgique are supporting the Bela non-profit organisation training programmes which introduce the functioning of the book sector to emerging authors. We will insert content on AI in them. For more experienced authors, we will have to create specific training courses. But, obviously, that takes a little time because we will first have to train the trainers.’  A project which, like the dozens of others cited, is vital to prevent seeing the book ecosystem succumbing to the shake-ups triggered by AI and crumbling.

Call for projects

A story, projects or an idea to share?

Suggest your content on kingkong.

send your ideas