Who Owns AI-Generated Content? And the Case of the Infamous Monkey Selfie
At first glance, the issue may seem straightforward. If you prompt the AI, then you own it, but this likely isn’t the case.
Why this is different than human-created content
To answer this question, you need to have a rough understanding of how AI learns to generate content. While traditionally, when we want a computer to complete a task, we will give it some input and then tell the computer how it should handle that input to get the output we need. When we use machine learning, however, we instead feed the computer an input and then show it what we want the output to look like. By repeating this process over and over again, we allow the computer to learn how to perform the task without us telling it exactly how it should be done.
To simplify this, let's say we give a computer the input of “tree” and then give it a picture of a tree. After only one round of learning, our AI certainly wouldn't know what part of the image was a tree, but if we give it many different pictures of trees, it will begin to pick up certain patterns until, eventually, it has a pretty solid understanding of what a tree is or at least what it looks like.
This is by no means a perfect understanding of how these algorithms work, but it explains a crucial part of the process, which is training them using a large data set. In the case of image generation, this would be a large dataset of labeled images. In the case of a chatbot, this would be a large dataset of text. This process, in particular, is what makes the intellectual property debate so confusing.
Where the intellectual property nightmare begins
The argument for divided ownership
Some people may argue that AI being trained on other people's images or artwork is the same as an artist looking at other people's artwork for inspiration. In reality, however, there seem to be some key differences. Most notably, AI can at times come close to copying some of the images in its training set. This is one of the main reasons many people believe that the actual ownership of AI-generated content should be divided among the people who created all the data used to train the AI model.
The argument for ownership by prompt
There are also many people that believe the owner of AI content is the person who came up with the prompt that created that image. This is often the first possibility that comes to mind when faced with the question. It does actually make some sense too. If you've ever been tasked with prompting these AI algorithms, you've probably learned that it can be a difficult task. Often times getting the result you're looking for takes a few attempts with various degrees of specificity throughout your request.
The argument for AI ownership and the case of the Monkey Selfie
You may be wondering what on earth a monkey has to do with this debate. Funny enough a monkey may have helped set some of the only legal precedents we have in this debate. It all started when photographer David Slater went to Indonesia to photograph a type of monkey called celebes crested macaques. Slater hadn't just attempted to get a picture of these monkeys however, he had instead set his camera up in such a way that would facilitate a monkey taking a selfie of themself. Surprisingly he succeeded multiple times.
Unfortunately for Slater getting the monkey to take his selfie may have been easier than the multiple-year lawsuit that these pictures would bring on. This lawsuit would debate whether Slater or the monkey himself owned the rights to the selfie. This case would end up setting the precedent that copyright protections are only applied to humans. So far this standard has been held up by the courts as well as the copyright office.
The argument for ownership by development
Some people will also argue that the designers of the algorithms used to generate the content own the rights to what it generates. Often times this will only apply if they had purchased or in some way owned the rights to all the training data that was used. However, this is still very much a legal gray area.
The argument for no ownership
It's not uncommon for people to believe that none of the above categories should qualify as ownership and would therefore leave the content in the public domain.
Ethical considerations
While I don't believe there to be an obvious winner in this debate, I do feel for artists and other creatives who have concerns about their work being used to generate new pieces which may resemble their own. As a photographer myself I too have some concerns about this. The only situation I see that seems more straightforward is if you would classify as the owner under all the above categories. For example, If you created an algorithm that was trained on your own content and you prompted it. However, due to the fact that the content is still AI generated it's still unlikely you'd be able to copyright the content it had generated.
How Shutterstock has handled the situation
Undoubtedly, the popular microstock website Shutterstock noticed that AI-generated images would likely become extremely valuable for commercial purposes and designed a system for distributing the earnings generated by these custom AI-generated images among all the creators whose content was used to help train the model. Shutterstock has dubbed this their creator fund.
If you enjoyed this article, please let me know by leaving a clap or a comment! If you'd like to see more content like this, please check out my profile or leave a follow. Thanks!
