Copyrights and AI

Gepubliceerd op 8 februari 2023 om 15:00



A lot can be said about copyright related to AI. As far as I know, no lawsuits have been filed in the Netherlands, but they have in the US, where some artists have sued Stability AI, Midjourney and DeviantArt. In the UK the same threatens to happen against Stability AI by Getty Images. In the meantime I've delved into the subject matter a little deeper, starting with the help of YouTube lawyer LegalEagle who made an informative video about it.

No human, no right

To start with the easiest requirement: computers, and therefore AI, can never get copyrights themselves, because they are not a person. Copyrights are exclusively reserved for humans. Even monkeys that accidentally take selfies cannot claim rights, not even with the help of animal rights organizations, as was tried a few years ago.

Originality and expression

A second requirement concerns the copyrighted work: it must contain a certain degree of originality and expression. In the early days of photography it was debated  whether or not to freely print the work of photographers, who after all only depicted reality as it was, and who could claim  copyright on reality? By now it has long been clear that choosing the subject, the camera angle, the lighting and everything else involved in a good photograph, together form one's own artistic input. The photo is therefore protected by copyright. Here too: the camera does not receive copyrights. The patent on the technique of taking photos has long since expired, so the inventors of photography cannot claim anything either.

Case law

In the US some lawsuits have been filed against individual works generated with the help of AI. Lower courts ruled that the process of generating could not contain creativity and that those works could therefore not be 'original' enough for copyright - and that is strange: These judges did not look at the product, but only at the process. This is new when it comes to copyright, and will therefore be brought before higher courts.

A graphic novel that was first granted copyright based on its content, 'Zarya of the Dawn', had it  revoked a month later. The author must now first prove that she has had sufficient personal input in the creation of the work. It is clear that she came up with the storyline, put the pictures in the right order, and added the texts. The main issue here is whether she can also claim the rights to the individual pictures, firstly because she used the Midjourney AI for the generation, and secondly because she used the name of actress Zendaya in her prompts. Indeed, her main character looks suspiciously like Zendaya. To be continued….

The input


The subject matter above was all  about the output: the generated work. The second discussion is about the input. Consider Zendaya in the case above. As a well-known citizen of the world, she has portrait rights. This means that her face cannot be used in commercials or other published artistic works because her face is worth money. She must give permission first. Can her portrait be used to train AI? That is a gray area still. The same applies to all copyrighted works: should all right-holders and authors have the opportunity to opt out, so that their works cannot be used?


Of the three best-known text-to-image generators, DALL-E 2, Midjourney and Stability AI, only with the latter it is known exactly on which images it has been trained. These come from Open Source Data collector LAION, a German non-profit organization that wants to democratize the training of AI, so that Big Tech cannot gain too much power. LAION has indeed scoured the entire internet and created a database with text-image pairs. They did not store the images in their own databases, but only recorded the link to each image. Stability AI then normalized these images into usable training data by scaling everything to more or less the same size, 512x512 pixels by default. I won't go into the details of training here, but what is stored in a neural network is not the picture itself: it is the similarities and differences between the pictures. Just as humans "train" their brains by seeing many cats, both in real life and depicted in pictures, an AI also learns what a cat is by seeing many pictures of cats in combination with the word "cat". It learns to recognize all the characteristics of a cat, and can then reproduce it. Because only one Zendaya is really known on the internet, a neural network will store and reproduce all characteristics of only this Zendaya.

Should training with celebrities be banned? Or should one mainly adhere to existing rules when publishing (generated) pictures of celebrities?

Fair use

LAION and Stability AI are not the first to scour the internet to use the data. Google does the same! It not only reads the entire internet, but also publishes it back to the internet. Lawsuits have been filed in the US against showing those images (without paying rights over them), as well as against showing pieces of text. Google won both by invoking the 'fair use' principle. This 'fair use' is an unknown rule in Dutch law, it is simply not part of Dutch or European copyright law. 'Fair use' means that Google does not use the copyrighted works for their original purpose, but a new purpose, and also does not show the entire work, but only a thumbnail or a piece of the text. Under these (and a few smaller) conditions, Google may continue to exploit the Internet. It is very well possible that this reasoning will also apply to AI, but we will have to wait for the results of the various court cases for that.

Using other people's work


In any case, what the AI does not do is create collages from existing, copyrighted works - unless explicitly manipulated to do so by the human user. Dutch law is relatively clear about the extent to which it is permitted to draw freely from other people's work for collages: it may no longer be recognizable as such (or at a glance). The collage must form such a new work of art that the input of the original works of art is completely subordinate. Here too, the process of making is not up for discussion, but only the end result counts. When discussing the legality of a collage, a judge will decide on the copyright, but the judge will only do so by looking at the work itself, not at the process.

Digital collages

A form of collage that does not stand out, but is often used in digitally created art, is 'matte painting'. If an artist wants to render rocks, forests, or crowds as realistically as possible (and not draw all the details by hand), they can take a piece of an existing photo with the texture they want, rotate it, magnify it, double it, and just as long edit it until it fits seamlessly into the artwork. It is not possible to determine from the end result where that original piece of photo came from. Those who watch movies or play games will see matte painting in action without realizing it.

Wafer-thin border

Games and movies are also the two genres of art where the use of computers has taken hold the most. The algorithms that calculate textures, lighting and backgrounds are only a small step away from the next step: AI-generated images and AI-generated videos. Current games, animation films and special-effect scenes already make extensive use of intelligent algorithms, from expert systems that calculate physical forces and accelerations to pre-cooked motion simulators, where the artist only needs to adjust a few parameters. The line between these tools and the neural networks that are the subject of this blog is starting to get wafer-thin here.


My expectation for the copyright battle is that AI image generation, like taking photos and making collages, will be seen as a process with sufficient creative input from the artist themselves. The AI can be seen as a tool - in such a way that each individual work is subject to copyright - as long as the author himself makes sure that he does not commit illegal plagiarism himself. So don't make copies of existing works (not even by accident), and don't depict well-known citizens of the world. If you think I have done that in my own creations, please let me know, so I can correct myself.


For those who want to dig further into this topic, here are some links:

Reactie plaatsen


Er zijn geen reacties geplaatst.