The Double-Edged Sword of AI: How Bookwire Can Help Protect You and Benefit from New Opportunities
Bookwire Spain Report – AI Initiatives
AI is affecting the whole book industry and brings lots of new opportunities, but also challenges. While more and more publishers are exploring different ways to use AI to increase efficiency in internal processes – lately driven by the rise of AI agents – using AI technology is a double-edged sword for the creative industries. Developers of AI models have been making use of illegal content databases to obtain data for the training of their models free of charge and without the consent of the rights holders.
From our point of view, AI is here to stay. It has been widely adopted by many industries already, attracting billions of dollars of investment to enhance models and open up for additional use cases, disrupting business models at a fast pace. Companies making reasonable use of AI technologies will gain a competitive advantage. What is important, however, is how AI is used to protect the rights of creatives, preserve values and create real added value for all parties involved, including publishers, authors, narrators, etc.
Bookwire has developed an AI strategy balancing the use of AI technology where mutually benefitial while taking measures to protect the titles distributed on behalf of our clients the best way possible.
TDM Opt-Out
Article 4 of the EU directive 2019/790 allows the use of content for „reproductions and extractions of lawfully accessible works and other subject matter for the purposes of text and data mining“. Text and data mining is defined as analysis of content to identify patterns and trends. Most AI companies argue that training their models with copyright protected content without getting consent from rights holders or paying any form of compensation is covered by the EU directive.
However, the directive allows for rights holders to declare a machine readable reservation not giving permission to text and data mining for their works, the so-called TDM opt-out.
Bookwire has already implemented different TDM opt-out measures, automatically applied to all titles delivered by default, unless a publisher explicitly opting out. This includes a TDM opt-out notice in ONIX (only applicable with ONIX 3.1 or higher) for eBooks and audiobooks, and the implementation of a TDM opt-out notice using the TDM Reservation Protocol (TDMRep) developed by the W3C in all EPUBs. TDM opt-out notices included in the files delivered by publishers to Bookwire already are fully respected and not overwritten.
While a TDM opt-out notice in ONIX and EPUB provides basic protection, they do have certain limitations. The opt-out information is not publicly available, but only delivered to the retailers Bookwire is connected to. Third parties getting hold of the files might alter them, getting rid of metadata and/or TDM opt-out information included (e.g. social media platforms). Therefore, Bookwire is working on implementing the International Standard Content Code (ISCC) as an additional measure. The ISCC is a content-derived identifier for digital media assets (including encodings of text, images, audio, video) using cryptographic and similarity-preserving hashes to create a unique ID for each digital asset. Bookwire plans to create an ISCC for every title in our database and store them on a public registry with individual metadata for every asset, including TDM opt-out information. Because the TDM opt-out information is stored publicly and separately from the actual files, everyone planning to use a certain title can easily check, if the respective title is protected by a TDM opt-out or not. This is especially important in order to undermine the argument made by AI companies they did not know of any TDM opt-out applied for a specific title.
AI Licensing
Currently, the best way to protect the works of rights holders is to sign AI licensing deals though. In a bilateral agreement clear rights and obligations for the AI company can be defined, like transparency measures, reporting of usage and, of course, remuneration. Bookwire is in contact with different AI companies already, interested in licensing training data of various kinds from the publishers we are representing. Our goal is to develop a mutually beneficial framework, giving back control over their works to rights holders and adequately compensating them while providing high-quality content not legally publicly available to AI companies.
The licensing market is still in its early stages. There are no standard license agreements, yet, and licensing fees do vary a lot, if publicly known at all. We do expect an active licensing market to develop over the next 18 to 24 months though, shifting from big technology companies developing foundational generative AI models to smaller start-ups developing AI applications for specific use cases. This will increase our bargaining position, getting better deals for publishers to opt-in to, while allowing us to increase pressure on lawmakers to better protect creative works from being used by AI companies illegally.
Bookwire OS becoming a digital publishing agent
Of course, we are also exploring ways to enhance our product experience using AI technology, providing sophisticated features and valuable service to the publishers using Bookwire OS. There are many ways AI can be used ethically and sensibly to provide better analytics, automate workflows, reduce effort for publishers and increase sales. Stay tuned for more updates in the coming months.
The best is yet to come.