Books are currently being used at scale for AI training and other applications, often without permission and without compensation to authors or publishers. Bookpact.ai offers an answer to this.
The six principles of Bookpact.ai
The AI economy changes almost weekly, and legislation is still very much in flux. That is why we operate from six guiding principles:
1.
Prior consent
Authors and publishers have the right to determine in advance whether and how their work may be used by AI companies. Consent is granted per title, per AI client, per type of use, and for a specific duration.
2.
Fair compensation
Authors and publishers are entitled to fair financial compensation for the use of their work. The ability to say ‘no’ to specific agreements, now and in the future, is a prerequisite for this.
3.
Content integrity and attribution
AI applications typically produce an interpretation of a work’ content rather than an exact reproduction. Whether this is a summary, a visualisation, an interactive application, or otherwise: authors have the right to ensure that the content is correct and properly attributed.
4.
Traceability
Authors and publishers have the right to have the use of their work traced, measured, and clearly reported within every AI application, as well as in training data.
5.
European focus
Bookpact.ai is designed to be independent of any single technological platform and, wherever possible, always chooses open-source and European alternatives.
6.
Sustainability
AI is energy-intensive. Bookpact.ai prioritizes AI companies that transparently report and demonstrably limit their energy and climate impact.
How Bookpact.ai works
Bookpact.ai is a two-sided platform that publishers can join on one side and AI companies on the other. Within their own portal, publishers (with the consent of authors) specify which titles may be used and under what conditions. AI companies then submit offers for targeted licences on a per-title basis. Publishers either accept or decline these proposals. Upon acceptance, access to the material is granted with legal and technical security measures in place. Publishers receive client- and title-specific reports that can be used in the royalty statement for the author.
4 types of use by AI
Bookpact.ai distinguishes four types of use for books within AI applications. Each type encompasses many variants, and different licensing arrangements are possible within each. Given the pace of technological development, this will continue to evolve.
1. Reference
AI applications consult books to answer user queries, with correct attribution to the end user. The manuscript is not shared; only the relevant passages are consulted.
2. Transformation
The content of a book is converted into a new form: an AI-generated summary, audioversion, video, or interactive application. The manuscript is shared, but is protected with a watermark and may not be used as training data.
3. Training
The content of a book is used to train or fine-tune language models. The manuscript is shared with a watermark.
4. Public AI
The work is made available for academic research, non-profit applications, or Dutch and European AI initiatives that could use a boost.
Created by the book industry
Bookpact.ai is founded for and by the book industry: publishers and book trade professionals who understand the value of content, the importance of copyright and the responsibilities that come with both. We believe deeply that makers should be rewarded for their ideas and creations. That’s not marketing, it’s why Bookpact exists.
“The only way to keep creativity thriving in the age of AI is through fair attribution for makers. Bookpact.ai puts control back where it belongs: with the creator.”
– Sander Ruys, founder Bookpact.ai
Interested?
When your publisher joins Bookpact.ai, they will reach out to discuss your participation. If they have not joined Bookpact.ai yet, you can refer them to us.
If you have any other questions, you can contact our team.
