Publishers demand transparency and fair compensation for AI training data use, launching the 'Frankfurt Appeal' at the Frankfurt Book Fair.
Frankfurt Book Fair launches copyright appeal
At the 78th Frankfurt Book Fair, the Börsenverein des Deutschen Buchhandels, which represents 120 publishers, introduced the "Frankfurt Appeal." This appeal urges the European Commission, Parliament, and the German government to demand transparency about how protected works are used in training artificial intelligence systems.Appeal demands detailed data disclosure
The Frankfurt Appeal calls for AI providers to disclose "concrete training data for each work," enabling rights holders to identify specific books or texts used in AI training. This is more detailed than the current European AI Act, which only requires summaries of training data, not a full inventory of individual works.Call for fair compensation and licensing
The appeal also proposes the creation of a "licensing market with equitable revenue models" to ensure publishers are compensated for the use of their catalogs. This addresses the European exception for text and data mining, which permits certain reproductions of legally accessible works, provided rights holders have not explicitly forbidden it.Support from major publishing entities
The Frankfurt Appeal has received backing from major publishing groups such as Penguin Random House Verlagsgruppe, Holtzbrinck Buchverlage, and the Federation of European Publishers (FEP). Discussions at the fair continue on whether legal action should be taken before new AI regulations are finalized.Dutch publishers launch AI regulation platform
Dutch publishers have introduced Bookpact.ai, a commercial platform that allows AI companies to use books only with the publisher's consent. The platform operates on an opt-in principle, giving publishers more control over how their content is used.Japan develops AI text conversion tool
In Japan, the National Library of the Diet is developing NDLOCR-Lite, a free tool that converts images of Japanese books into text without needing a dedicated graphics processor. This aims to make historical collections more accessible while potentially aiding AI training.Australia opposes AI copyright changes
In Australia, 17 organizations representing publishing, media, and the arts have urged the government not to reopen copyright laws to benefit AI companies. They argue that the use of protected works should require negotiated licenses, not automatic exceptions.China warns against AI use in books
Huaxia Publishing House in China has started including notices in some books stating that their content cannot be used to train AI. However, enforcing this restriction technically remains a challenge, as it relies on compliance rather than technical barriers.Global concerns about AI and copyright
Across the world, concerns about AI's impact on copyright, transparency, and content quality are growing. In the U.S., a survey of 559 book industry professionals found that AI is being used but with worries about legal and ethical issues. In France, a researcher showed that Kindle e-readers could be vulnerable to attacks through malicious websites, raising security concerns for connected devices.