How much AI is acceptable?

A book has a title, an ISBN, a publishing house and the name of a human being on the cover. Children’s books also name the illustrators, translated literature the translators. Pieces of music list composers, lyricists, performers and producers. Films also carry extensive credits. We hardly ever learn how the content was created and what share artificial intelligence has in it. Journalism, too, does not comprehensively declare AI use – sometimes for the understandable reason that practice is changing within a short time and AI is taking over an increasing share in the production of content.

Generative AI is increasingly affecting the core of both production and product, firstly because of the volume of texts, images, music, voices and videos that can now be produced with AI, and secondly because AI-generated content creates the appearance of human creativity.

The audience wants clarity

If you ask the people for whom content is ultimately created, you get a fairly clear answer.

• Books. A survey conducted by Books on Demand (BoD) and Appinio in September 2026 among 1,000 people in Germany found that 49% accept books in which AI has been used in a supporting role, and 30% even fully AI-generated books. At the same time, 83% are in favor of labeling AI use. (https://www.bod.de/pressemitteilungen/umfrage-ueber-80-prozent-der-leserinnen-wuenschen-sich-ki-kennzeichnung-von-buechern)
• Music. The international survey by Deezer and Ipsos from November 2025 covered 9,000 people in eight countries. 80% want a clear label for fully AI-generated music. 73% of streaming users want to know whether a platform is recommending AI music to them. In a listening test, 97% did not assign all three pieces played correctly. (https://newsroom-deezer.com/2025/11/deezer-ipsos-survey-ai-music/)
• Content of all kinds. According to a study by the consulting firm Baringa from 2025 based on international surveys of more than 5,000 people, 77% want to know whether content has been created wholly or partly with AI. 62% support visible identifiers or watermarks for AI use. (https://www.baringa.com/en/insights/balancing-human-tech-ai/trust/)

Even among those who accept AI-generated content, there is an interest in information about how it was produced.

Most cultural creators and producers support AI transparency

Many creative professionals, their professional associations and collecting societies call for transparency. Their statements concern both the labeling of new content and the use of protected works by AI providers:

• CISAC and IFRRO. The international umbrella organization of authors’ societies, CISAC, demands consent, royalties and transparency for AI uses. It calls for both the disclosure of the works used for training and the recognizability of AI-generated results. The international federation of reprographic rights organizations, IFRRO, emphasizes the documented use of texts and images, rights clearance and licensing. (https://www.cisac.org/AI; https://ifrro.org/resources/documents/General/V2_IFRRO_AI_Position_Statement_29_March_2023.pdf)
• European associations. The European Writers’ Council (EWC), the translators’ association CEATL and the Federation of European Publishers (FEP) call for the labeling of AI-generated books. The EWC supports human- and machine-readable information that is retained along the value chain. (https://europeanwriterscouncil.eu/2504_labellingai/)
• Swiss professional associations. The association Autorinnen und Autoren der Schweiz (A*dS) calls for the labeling of machine pre-translated literature. The journalists’ organization impressum demands disclosure of media contributions that are fully or substantially generated with AI. The association Filmregie und Drehbuch Schweiz (ARF/FDS) calls for responsible, transparent AI use and, in particular for documentary images and sounds, for labeling of significant interventions. (https://www.a-d-s.ch/uploads/media/default/451/240305_Positionspapier%20%C3%9Cbersetzung%20und%20KI_deutsch.pdf; user_upload/Resolution_September_2023_Kunstliche_Intelligenz.pdf">https://www.impressum.ch/fileadmin/user_upload/Resolution_September_2023_Kunstliche_Intelligenz.pdf; https://arf-fds.ch/de/downloads/verband/arf-fds_ki-positionspapier_09.2025_de.pdf)
• Visarte and SONART. Visarte deals with the consequences of generative AI for the visual arts and the protection of creative services. SONART, the association of music creators, is particularly committed to copyright protection, fair royalties and the conscious use of AI tools. For both, rights in works used and the economic position of human creatives are key concerns. (https://visarte.ch/wp-content/uploads/2023/09/Stellungnahme_Visarte_Kulturbotschaft-2025-2028_def._d_21.9.23-1.pdf; https://www.ige.ch/de/blog/blog-artikel/plattformen-vermitteln-den-eindruck-musik-sei-ein-kostenloses-gut)

Publishing houses, producers and industry organizations of the creative industries are also acting in a forward-looking way. On the one hand, they use artificial intelligence to simplify and improve their production. On the other hand, they are aware that their own role is changing and being called into question in parts. Current initiatives emphasize human authenticity – they are not yet widely established.

• “Without AI”. Since February 2026, the German children’s and young adult publishing house Loewe has been marking its available books with “Without AI”. The publishing house thus declares that texts and illustrations come from humans and backs up this statement through agreements, review processes and editing. (https://www.boersenblatt.net/news/verlage-news/loewe-fuehrt-das-neue-label-ohne-ki-ein-412729)
• “By humans”. At the beginning of October 2026, 18 children’s and young adult publishing houses introduced a joint identifier “By humans · written · designed · reviewed”. Participants include Arena, dtv junior, KOSMOS, Moritz, Ravensburger and the Swiss NordSüd Verlag. The agreement concerns human creative services and editorial review; supporting AI in administrative processes remains possible. (https://avj-online.de/themen/von-menschen/)
• “100% author · without AI”. Since September 2026, Books on Demand has also been making the identifiers “100% author · without AI” and “AI-assisted” available to other publishing houses. These are voluntary self-declarations; BoD does not verify their accuracy. (https://www.bod.de/ki-label)
• “Création humaine”. In France, Librinova offers the “Création humaine” label. Certification combines contractual declarations, an automated text review, a conversation with the person who writes and random checks of the manuscript. (https://www.librinova.com/offre/services-a-la-carte/publication/label-creation-humaine-texte)
• “Human Authored”. The American Authors Guild has set up “Human Authored” with a public register that has been open since March 2026 to all authors with US publications. The British Society of Authors launched a similar system in the same month; by September, more than 2,000 books had been recorded there. Both programs rely on authors’ declarations. (https://authorsguild.org/human-authored/; https://humanauthored.co.uk/six-months/)
• PHOTOLABEL from Switzerland. PHOTOLABEL, an association founded in Bern in summer 2026, offers a positive identifier for authentic photographs. Its rules for capturing and editing are based on photographic practice and World Press Photo. The license for the identifier is tied to defined requirements for each individual photograph. (https://www.photolabel.org/intro-d; https://www.photolabel.org/guidelines-d)
• IFPI. In July 2026, IFPI and other music associations introduced the voluntary identifiers “AI-Generated” and “AI-Assisted” for sound recordings. The information is to be passed on to music platforms via metadata. (https://www.ifpi.org/music-community-introduces-new-labelling-program-to-distinguish-generative-ai-in-sound-recordings/)
• “Human Provenance in Film”. The “Human Provenance in Film” initiative is developing a standardized declaration on AI involvement in completed film and television productions. The draft distinguishes three classes and provides that the declaration can be passed along with the respective version through production, financing, distribution and licensing. The consultation runs until the end of October 2026. (https://humanprovenance.film/)
• Press Council. The Swiss Press Council recommends transparency on AI use and demands full journalistic responsibility. (https://presserat.ch/journalistenkodex/ki_leitfaden)

How sensitive the question of origin has become is illustrated by the current literary dispute surrounding Thélyson Orélien. The author of the successful novel “C’était ça ou mourir” was accused of using AI. He denies that AI wrote the novel text and separately admitted to plagiarism in earlier texts. The case was discussed intensively in Western Switzerland. At the same time, it shows why any suspicion and AI detectors must be handled carefully. ProLitteris was consulted by Le Temps for the linked contributions. (https://www.letemps.ch/culture/livres/thelyson-orelien-dement-avoir-utilise-l-ia-mais-reconnait-des-plagiats; https://www.letemps.ch/culture/livres/la-polemique-ne-va-pas-m-enlever-mon-emotion-les-librairies-romandes-continuent-de-soutenir-thelyson-orelien)

Bookstores, too, can hardly verify how each title in their full range was created. Orell Füssli and the book retailer Lüthy told SRF that they support transparent information from publishing houses. This is relevant for the immediate customer relationship: depending on the circumstances, an explicit but incorrect assurance about origin can have consequences under sales law or unfair competition law. The Börsenverein des Deutschen Buchhandels recommends integrating information about AI use into product metadata via the ONIX standard. ProLitteris also spoke in the background with the editors for the SRF feature. (https://www.srf.ch/sendungen/kassensturz-espresso/kassensturz/literaturbranche-im-wandel-ki-generierte-buecher-das-muessen-sie-wissen; https://www.boersenverein.de/interessengruppen/ig-produktmetadaten/best-practices/)

Why label? Originality, trust asset and fairness

There are voices that are indifferent to AI. Why label at all? Does it not simply come down to whether someone likes content? The discussion at the recent Frankfurt Book Fair shows why information on how a work came into being is important. 

We at ProLitteris see three good reasons for this and one compelling reason.

• The first good reason: originality. Many people want to read, listen to music and view images created by real humans. They are interested in personal experiences, thoughts, imagination and forms of expression. In relationship-based communication, the person behind content plays an important role.
• The second good reason: trust asset. Reading time and attention are investments. We notice whether a book is interesting when we read it. The way it is produced thus becomes a trust feature that parts of the readership want to see. This need becomes more acute in the case of photographs and genuine sound and image recordings. In journalism and science, authenticity carries particularly high weight.
• The third good reason: fairness. Humans like humans and live in human relationships. Many consumers therefore wish greater success for a human creator than for a person who presses buttons and masters software. There are readers, listeners and viewers who want to acknowledge and support human services.
• And the compelling reason: copyright! Anyone who does not directly or indirectly ensure and express that their content constitutes protected works or performances under the Copyright Act (URG) or a similar act of another legal system calls their products into question as assets. This is because pure AI output is not protected. What matters is the human creativity of individual elements of a piece of music, a literary or visual work, an audiovisual file or a computer program. The copyright threshold is not high; it is often lower than quality requirements derived from other, non-legal considerations. However, this threshold should not be undercut if you take your content seriously. As regards a collecting society such as ProLitteris, SUISA, SSA, Suissimage and SWISSPERFORM, rights holders are in any case obliged to claim royalties only for protected objects.

The European AI Act also prescribes transparency about AI use, in particular machine-readable labeling of AI-generated content and, in rarer cases, the disclosure of deepfakes and similar modifications of genuine depictions and recordings.

For all these reasons, labeling can be both a legal measure and a strategic decision. Designing a policy starts in the production process. Publishing houses, studios, editorial offices and platforms can define which creative services they reserve for humans and which AI applications they allow. This policy can be formulated as a recommendation to authors and other creative professionals who contribute.

Such a recommendation, here using the example of an academic journal, reads as follows:

Recommendation to authors on the use of AI

Our journal values independent specialist contributions that reflect the individuality, professional expertise and personal style of their authors. This is also intended to maintain the prerequisites for copyright protection as an individual human creation.

We therefore recommend that our authors personally develop and shape the defining elements of a specialist contribution, in particular:

• Concept and train of thought
• Main messages and conclusions
• Language and personal style
• Selection, arrangement and coordination of statements, facts, opinions and sources

AI may be used for research, quality assurance and language support, but the text and its essential parts must not be handed over. Academic and linguistic design remains the responsibility of the authorship.

We also recommend that authors know and consider the specific features and limitations of generative AI when using it. These include the proneness to error, the reliability of sources and evidence, the lack of transparency of the results and the protection of confidential information.

Responsibility for the entire specialist contribution, its statements and its sources remains with the authors. Copyright protection is a general condition for the management of rights in works and/or performances.

Only then does the question of disclosure and content labels arise, such as those offered by ProLitteris under MyCopyright for members.

Three connections to copyright

We have shown that there are various reasons to support transparency, and there is actually little to be said against it. The copyright dimension is somewhat broader than mentioned earlier, and to understand it you have to look at the problems created by artificial intelligence. 

1. Use of protected works as raw material. Current and older protected works are used for training and further development of models, for retrieval systems and to generate new content. In many cases, rights holders do not know which files have been used for which purposes. Only if content and rights can be reliably identified can reservations of rights, consents and licenses be explained with greater precision. The European Commission is examining a register for reservations in text and data mining; the EUIPO is examining a networked directory with work information under the name CopyrightView. (https://digital-strategy.ec.europa.eu/en/library/new-feasibility-study-introducing-eu-level-registry-text-and-data-mining-opt-out; https://euipo.europa.eu/tunnel-web/secure/webdav/guest/document_library/observatory/documents/reports/2026_Mapping_EU_DB_metadata/2026_Mapping_EU_DB_metadata_ExSum_en.pdf)

2. Recognizability of protection status. Under Article 2 URG, Swiss copyright protects intellectual creations with an individual character. Foreign laws have similar rules. Anyone named on a copy of a work or at its publication is deemed, under Article 8 URG, to be the author until proven otherwise. For AI-assisted content, what matters is which defining elements can be traced back to human personal services. Concept, train of thought, linguistic or visual expression, selection, arrangement and the key conclusions can be relevant. Alongside the result, the creative process gains importance.

3. Human creativity as a distinguishing feature. AI-generated content competes with human services for attention, purchasing decisions and returns. Labeling can make human creativity visible and help to meet the challenges of artificial intelligence in the creative industries. It can be in the personal and economic interest of rights holders that pure AI products are recognizable and that human works remain recognizable.

To address these three challenges and to shape the future of rights management, ProLitteris has been working since the beginning of 2026 on an identification and trust infrastructure for content of all kinds.

MyCopyright: our offering for credible declarations

Together with Swiss Trust Layer and other technology partners, ProLitteris is developing the Swiss Copyright Office as an identification and trust infrastructure. We want to complement existing labeling initiatives and offer them a shared technical foundation. A declaration or assertion alone has little value if it is not received by an independent body and can be verified. (https://www.prolitteris.ch/en/about-prolitteris/services/mycopyright)

Evidence preservation is already available for specific files and versions using a cryptographic fingerprint, timestamp and the option of encrypted deposit. Through qualified identification, a declaration can be assigned to a specific natural person.

We offer differentiated declarations on the use of generative AI, on human authorship and on assuming responsibility after human review. A publishing house could declare the text, translation and illustrations of a book separately; comparable applications are conceivable for music, images, voices and film productions. Integration into partners’ processes and transmission via interfaces and international standards such as ISCC and C2PA are in place and are being continuously expanded.

As a further stage, standards and registers are planned for works, the people involved and rights, supplemented by machine-readable reservations of rights and licensing options. The infrastructure is to be organized independently and, with the involvement of collecting societies, made accessible to rights holders and partners in Switzerland and internationally.

The labeling initiatives mentioned in this article show that creative professionals, producers and platforms are developing their own responses. They become stronger when evidence, independent documentation and a shared data source are added.

With the Swiss Copyright Office, we are building the technical and institutional foundation that can connect such solutions and make their statements more traceable. How much AI is in it should be attributable to specific content and to the humans who stand behind it.