Kobo and AI: Making Reading and Publishing Even Better, Together
We know that artificial intelligence (AI) is a hot topic these days, and we want to be transparent about how we’re using algorithms and machine learning at Rakuten Kobo and on the Kobo Writing Life platform. Your trust is paramount to us.
Putting AI in the same sentence as authors, books or publishers makes alarm bells go off, and for good reason. The use of material under copyright for training is an ethical and legal issue in the creation of large language models (LLMs) and anyone who makes or sells any kind of intellectual property has every reason to be wary.
At Kobo, we have two non-negotiables when it comes to how we leverage AI. They are:
1) We don’t feed books into any LLM
2) We don’t use books to train any LLMs
Our approach is built on a simple promise: to help you connect with more readers while protecting your work. We’re committed to treating you fairly, safeguarding your intellectual property, and giving you the tools you need to thrive on our platform.
Our Commitment: Your Success is Our Success
We want to be crystal clear: Kobo does not use your books to train generative systems. We will never feed your content into anything that creates new works. Your books are your livelihood, and we’re dedicated to upholding the integrity of the publishing ecosystem.
We know that the jargon can be confusing, so let’s simplify things. We use tools that are AI or AI-adjacent, like algorithms, “really big algorithms”, and various machine learning techniques, all designed to make your life easier, help make the jobs of Kobo team members easier and connect you with more readers.
Some Key Examples:
• Protecting Our Catalog: We use machine learning models that we have developed ourselves to identify and quarantine books containing hate speech, child sexual abuse, or other harmful content that violates our guidelines. This protects your books from being associated with inappropriate material and helps keep Kobo a safe and respectful place for everyone. This model assists our human review team. It cannot be used to generate new content and is not shared with outside companies.
• Boosting Discoverability: We will use algorithms and machine learning to analyze samples of book text to improve categorization, search, browsing and personalization, so readers can find your books more easily. If you’ve written a story about “vampire cowboys,” we want to make sure the right readers can find it. This helps us fix any issues with inaccurate or incomplete metadata and gives us better insight into who might like to see your work. Rest assured, this process only uses samples of your text – we never share your entire book.
• Protecting Authors’ Intellectual Property: AI is creating a huge challenge for companies like us that run services for indie authors and publishers. Bad actors are using AI to create pirated, reprocessed works, summaries, internet content packaged as books, and other content that takes up space in search results and crowds out legitimate books. It also consumes time from our operations team, trying to detect and remove this content. As the models they are using become more sophisticated, we will need to use machine learning tools to find patterns and similarities to legitimate books we already have. We have always screened out books that were exactly like other books using much cruder algorithms, but as they get more sophisticated, we need to keep up.
These forthcoming updates reflect our commitment to using technology to improve your experience, protect your work, and connect you with more readers. Thank you for being part of the Kobo community and we're excited to continue this journey with you.
Best,
The Kobo Team Writing Life Team