Türkiye raises alarm over AI companies’ access to book archives

Claims that global artificial intelligence companies are turning to the archives of Turkish publishers to train their language models have sparked a new debate in the publishing industry.

As part of a special report titled “The AI Companies’ Hunt for Books,” the steps reportedly being taken by global AI companies to access book archives in Türkiye, following similar moves in Europe, will be examined through discussions with publishers and industry professionals, CE Report quotes Anadolu Agency.

Nevzat Argun, editor-in-chief of Nobel Academic Publishing, told Anadolu Agency that AI companies are attempting to collect extensive academic databases through intermediary firms, warning that the process poses a serious threat to both publishing and the production of scientific knowledge.

Pointing to growing interest in Turkish-language sources for training AI models, Argun said his company had also received a similar request.

He said that, through a Netherlands-based company, more than 3,000 books belonging to Nobel Academic Publishing and another 2,000–3,000 books from other publishers were requested.

“When we noticed that individual copies were being ordered, we became suspicious and uncomfortable with the situation. After our assessment made it clear that the request was connected to AI training, we canceled all the orders. We were among those who took one of the steps that started this debate in the publishing community,” he said.

Argun stressed that the process is not limited to purchasing printed books. He said offers had also been made to digital platforms that license publishers’ books as e-books, with requests to purchase all of the digital content available on those platforms in bulk.

Warnings over monopolization and a colonial approach

Argun also addressed claims that rare works on the market are being scanned and digitized before being physically destroyed or removed from public access, describing the developments as an attempt at global monopolization.

Argun argued that intermediary companies do not change the underlying structure, regardless of which country they operate from, and claimed that Western-based global capital is seeking to establish a monopoly over academic publishing.

“There is an imperial background in Western civilization that has become visible through the desire to demand more than one is entitled to. We are facing a similar colonial approach in publishing. They could have books collected one by one through an intermediary in Türkiye, scanned, and uploaded into their own systems. This risk remains on the table,” he said.

“Not open access, but a source-less narrowing of knowledge”

Argun said that content transferred without adequate controls into AI databases could damage academic traditions. He emphasized the difference between open access and uploading content into AI models.

“Academic publishing is built on accountability, peer-review processes, citations and bibliographies. With open access, researchers can reach a work, cite it, and the author’s contribution remains visible. But with AI models, content becomes mixed with other content; everything is blended together into a body of information whose source, author and peer-review process are unclear. If the source of original knowledge is not identified and its value is not protected, academic work itself loses its meaning. At a time when book sales have hit rock bottom, this process could discourage researchers from producing new work and ultimately impoverish knowledge production,” he said.

Translation memory is also being targeted

Argun said global companies are examining not only copyrighted domestic content but also the translation databases that publishers have built up over many years. He recalled a specific approach made around three years ago.

“A foreign publisher with whom we worked extensively and whose books we translated into Turkish made us an offer. They said, ‘If you give us the data from our books translated into Turkish, we want to enrich our AI; we will pay you in return.’ They later abandoned the idea. Most likely, they wanted to use AI to quickly translate the books in their own portfolio into the world’s languages and make them available as e-books. They already had the original texts. What they needed was our translation memory,” he said.

“A fair model is essential”

Argun stressed that AI technology is an inevitable part of technological development, but said a fair copyright and revenue-sharing ecosystem must be established.

He noted that copyright agreements and legislation have not yet fully adapted to the speed of technological development.

Argun also emphasized that if a domestic or ethical AI model is developed in the future, the rights of copyright holders must be protected.

“It is not technologically difficult to establish such systems. For example, if a database containing 100,000 books is created, publishers and authors should receive a share based on quotations from their works or the extent to which their works are used in AI-generated responses. Just as digital video or search-engine platforms distribute revenue to content creators according to viewing rates, a fair mechanism should be established in which the greater the contribution, the greater the share received,” he concluded.

Photo: Chat GPT