AI Policy

In Short

• Your book is translated by AI models. No human translator works on it, and nobody on our team needs to read it for the translation to be made.
• Your book is never used to train an AI model. We do not train models ourselves, and we only use AI providers that do not train on what we send them.
• Your book is sent to the AI in separate passages that are spread across several providers, so it is common that no single provider receives the whole book.
• You keep every right to your book and to its translation. We take no royalties, and you never have to credit us.
• Your files are deleted automatically within 90 days, and you can delete them sooner yourself.

What the AI Does

Each step of a translation is carried out by an AI model with its own job:

• Reading your book. Before translating, a model reads your book and prepares a translation guide, which keeps names, terms and tone consistent from one chapter to the next. You can review and edit it.
• Translating. A model translates the book one passage at a time, following the translation guide and any instructions you gave us.
• Checking the result. Other models check the quality of the translation and fix what they find.
• Reading scanned pages. When a PDF holds pictures of pages rather than text, a text recognition (OCR) service first reads the text from them.
• Tools you choose to use. If you edit your translation with our AI editor, a model reads and changes the passages you ask about, and it can search the web to check a reference. If you translate your book cover, an image model draws the new cover picture.

We choose the model for each step and replace it when a better one appears. We currently use models from companies such as DeepSeek, Google, OpenAI and MiniMax.

Where Your Book Goes

Before anything reaches an AI model, we split your book into passages. Each passage is sent as a separate request through one of these services:

• OpenRouter passes each request to one of the companies that run the model, which we call providers. We instruct OpenRouter to only use providers that, according to OpenRouter's data policies, do not collect or train on the text they receive. OpenRouter itself does not keep the text of our requests.
• Microsoft Foundry, Microsoft's AI platform on Azure, runs some of the models itself. Microsoft commits that the text we send is not used to train AI models and is not shared with the companies that made the models. Microsoft may keep a request for a limited time and have it checked when its automated systems suspect abuse of its service, and it does not use that copy for training either.
• Datalab reads scanned pages and keeps nothing once it has returned the text.

Because the requests are spread across providers, it is common that no single provider receives your whole book. We cannot promise it for every book, since a short book fits in a handful of requests, which is why the rules on this page apply to every provider we use.

These services receive the passages only to send a result back to us. We do not send your email address or any of your account details with them.

No Training on Your Book

We do not train or fine-tune any AI model on your books or on their translations. We have no model of our own to train, and we do not sell, license or give your book to anyone who builds AI models.

The providers we use do not train on your book either. OpenRouter only routes our requests to providers that its data policies list as not collecting the text, and Microsoft's terms rule out training. We rely on the policies that each of these companies publishes, and we choose our providers by them.

To improve the service, we sometimes run parts of a book through our system again, for example to compare two AI models or to check that a fix works. These runs use the same providers under the same rules. We may also send a few short passages, never a whole book, to specialised services that measure translation quality or detect AI-written text, and these services do not train on them either.

Who Reads Your Book

The translation is made entirely by our AI pipeline, so nobody on our team needs to read your book to translate it. We may read parts of a book or of its translation when you ask us for help with it, when we investigate a problem, or when we study the quality of our translations. You can also invite a proofreader to your translation, who can then read it and mark its errors.

Your Rights

You keep all rights to your original book, and we claim no ownership, copyright or other right over it or over the translation we produce. We take no royalties or share of your sales, we place no limits on what you do with the translation, and you never have to credit us. The Terms of Service set this out in full.

You do need the right to translate the book you upload. If you are not the rights holder, for example when someone else publishes the book, you must have their permission first.

How We Use AI Ourselves

Beyond translating books, we use AI assistants to help us write the software behind the service and some of the text on this website. When we investigate a problem with a translation or work on the quality of our service, these assistants may see parts of a book. They are professional AI tools, used under terms that rule out training AI models on the content we give them, so your book is not used for training there either.

Changes to This Policy

We update this page whenever our use of AI changes, for example when we start using a new AI service, and we change the date at the top when we do.

Questions

If you have a question about how we use AI, or would like your files deleted, please write to us through the contact page.