When last October’s “Artificial Intelligence (AI) created work” Edmond de Belamy “was auctioned for about 380,000 euros at Christie’s, this posed new questions to the art world. Who is the author of such works and who owns the rights to them? In an interview, the lawyer Dr. Patrick Ehinger the law and says in which other areas there is still a need for action.
LEAD: Mr. Ehinger, what are the legal issues with pictures like “Edmond de Belamy” made by an AI?
Patrick Ehinger: AI is usually associated with use for analytical purposes, such as image recognition or the analysis of consumer e-commerce behavior. Here, on the other hand, a creative work was created by means of AI. Therefore, one can also speak well of “artificial creativity”.
The legal problem is that the prerequisite for copyright protection is always a human creation process. If you now have software that creates a work virtually autonomously, it is very doubtful whether such a human creation process actually still exists.
LEAD: But the AI software was developed by one person.
Ehinger: That’s right. But you have to differentiate here. The one question is: who is the author of the AI software? And the other: What about the products of this software? Who is the author of this? Basically, the developer is only the author of the software and the user is the author of the works created with the software.
Of course, there are exceptions to this, for example when the developer has already planned and anticipated how the product actually looks in the end when programming. That software is created for this purpose should be the exception. In this case, the developer is usually the user as well.
LEAD: When would the criterion of the human creative process be fulfilled?
Ehinger: If the software in the overall view is a technical tool for humans. For example, the photographer also uses a camera or later the software Photoshop as an aid without losing his copyright status. As far as we know, in the case of “Edmond de Belamy” the image was created almost completely by an AI software. The artist group Obvious has also not developed the software itself, but only trained with data. Of course, one can then argue about whether this “training” of the software is enough to justify a human creation process. To do this, one would need to look closely at the training data and see if it was possible to anticipate the end product sufficiently.
You might also be interested in this: When digital art becomes boundless experience
LEAD: How would the users have to act in a specific case so that their work would be protected by copyright?
Ehinger: One possibility would be to give the software concrete guidelines for the design of the work. For example, in one image, you might decide in advance that the result should be a still life, with two red roses in a blue-striped vase and a certain yellow tint as the background.
The more such presets you make, the more clearly you have anticipated the picture and the better you could argue that it is a human creation process. However, it is also conceivable that the software makes suggestions, from which the user then makes selection decisions and in this way designs the product. For example, you press a button, the software makes a suggestion for a picture and you say: Ok, I do not like the eyes and the hair in the portrait, we do it differently.
The more suggestions you get and the more changes you make, the more the software becomes a technical tool. But there are also many other possibilities for the user to make such design decisions. It would also be conceivable, for example, to link to the preselection of the data that the KI software accesses and from which it generates the image.
LEAD: Will such legal borderline cases become more frequent as a result of digitization?
Ehinger: That will show. Digitization is about to cover virtually all areas of the economy and working life. This increasingly includes the creative industries. For example, there is more and more music that was composed with KI software. AI is also used in the design of computer games. Of course, this trend will increase as the better this “artificial creativity” works.
LEAD: So what should users be aware of when using AI software?
Ehinger: If you license and use a software, of course you want to be the only one who is allowed to use the products. For example, other people should not be allowed to copy and make these products public without my consent. Ultimately, the question is always the same: is the software that I use in the concrete case still an aid or does it work so far autonomous that one can not say anymore, here is a human-controlled creation process?
So who wants to have copyrights, should use a software in which you have as many ways of influencing and making design decisions. It would be good if the software documents the individual steps as far as possible.
LEAD: What other legal means are there in case copyright does not work?
Ehinger: In connection with AI products, in practice, without any doubt, the ancillary copyright will be relevant. For the ancillary copyrights do not have the goal, like copyrights, of protecting the personal intellectual creations of man, but also include those which are worthy of protection for other reasons, such as the investments required for this purpose. This also applies to cases in which the human contribution to the creative process is too small for copyright to take effect. Examples are the protection of photographs or databases.
LEAD: What could be the future legal regulation that includes the works produced by KI?
Ehinger: You have to wait for that. Very different positions are currently represented and the range between the individual solution proposals goes very far. In my opinion, it would make sense first to ensure that the training results, which are partly calculated with a high financial outlay, are protected by KI software. However, this concerns less the products of artificial creativity, but rather the development of the used software itself.
You might also be interested in this: Google Arts & Culture opens digital 3D archive