AI and Copyright

You can’t share everything with an AI model.

There’s some important rules to follow, when using others’ content in your AI prompts or when training your own AI models.

RettighedsAlliancen have developed a guide that gives you a quick overview of how copyrighted content can legally be used to train AI.

This includes the use of books when training artificial intelligence, uploading materials to GPTs, etc.

Content from publicly accessible websites from before June 7th, 2023

Publicly available content collected on websites before June 7th, 2023, may not be copied or used to train AI models, unless explicit consent has been given by the relevant content rights holders.
Prompting with copies of copyrighted content

It is not allowed to prompt an AI with copies of copyrighted content, e.g. a book chapter, as a digital copy of the text is created by the AI servers.
Uploading copies of copyrighted content

It is not allowed to upload copies of copyrighted content to an AI service without prior permission from the relevant rights holders. For example, uploading a book to serve as 'supplementary knowledge' when creating a GPT with OpenAI's GPT service. In this example, digital copies of the text are created on OpenAI servers. In addition, copies of the books are made available to the public if the GPT is published.
Consequences of non-compliance

Non-compliance with copyright laws when developing AI models and services has consequences. The rights holders of the protected content can demand that the models or services be removed from the internet and destroyed. They may also demand compensation and remuneration for unauthorized use of their content.