The AI art generator trend has sparked huge controversy in the art world. Some see AI art a threat because they believe AI art will take over human artists. While nobody can deny that artificially generated images do affect artists in some way, there is still the copyright issue. When the Lensa AI art app recently went viral, many found left over signatures from the original artworks Lensa uses to make new art. So is AI art stealing from artists? Let’s get into it in this blog post!
When AI artwork started flooding the internet in 2022, I wasn’t interested at all. I thought it was just the next new craze that’s going to ebb soon – like so many other things. If you know me, you know that I’m not about ‘the next shiny thing’. I like to stay focused on creating art and believe that running after the newest shiny object only deters you from your creative process.
But the AI art generator trend seemed to stay and I saw more and more digital artists jumping on the bandwagon. So I decided to look a bit more into it. First, I watched some AI art tutorial on YouTube. Some guy demonstrated how you can generate really impressive images by just feeding the machine a couple of words of text. But then he had the system create the Taj Mahal in Van Gogh’s ‘Starry Night’ style. I was like, mmh… how do I feel about that? Not so cool.
This was the first time that I felt a bit uneasy about AI art, because in that instant it was like AI art is stealing from artists.
Long story short, I set out to try AI art myself and see what it’s all about. After having signed up for the free Midjourney trial, I simply typed in some prompts describing a random image I had in mind. The result came super-fast and it looked pretty amazing, considering that this was my first attempt at creating AI art. I actually used that AI-generated image as background in this surreal artwork of mine.
But already when I went on creating my second piece, things got a bit weird… I noticed that one of the images had some funny sort of letters in the bottom corner. They could easily be the remains of a signature or marketing slogan. Wonder where that’s coming from… Does that make AI art ethical?
In this image generated with Midjourney I found the remains of some sort of lettering in the bottom right corner. Looks like a signature or a slogan. I wonder where that is coming from…
After looking deeper into how AI generates art from text, it is not that astonishing anymore. I found many online articles about artists claiming that AI art is theft and that artificial intelligence has been exploiting their artwork.
It is important to address the AI art controversy because it affects us all, especially as artists. That’s why I started a series of posts that deal with the AI Art vs. Human Art debate. I am trying to look at the topic from different angles, share opinions of other artists as well as my own views. I’d like to know what you are thinking, so join in the discussion and leave your thoughts in the comments!
In this post, we want to delve into the topic of whether AI art is stealing from artists. So let's get into it!
Is AI art stealing from artists?
While text-to-image generation software has been around for a number of years already, it only went mainstream in 2022 with programs like Midjourney, Stable Diffusion, DALL-E, and most recently Lensa. They are finally “bringing art to the masses,” as the developers like to put it. In fact, it is very easy for anyone to create impressive-looking imagery by just typing in a few descriptive text prompts.
For the above image that I created with Midjourney, for example, I only had to type in something like “dreamlike jungle with big vibrant flowers,” and there it was.
The debate around whether AI art is stealing from artists flared up again when artists spotted their original signatures in Lensa’s AI art. Lensa is an iOS photo editing app that works as follows: users upload their photos and artificial intelligence generates portraits in different art styles.
Dozens of images generated with the AI art app Lensa show signatures of the original artists whose works the app drew from to put together new creations. (Image via @tinymediaempire on Twitter)
To make these seemingly unique AI art portraits, Lensa relies on Stable Diffusion, an image synthesis model (ISM) trained using the online database LAION-5B. Exactly this ‘training process’ is the reason for the AI art controversy. LAION-5B pulls publicly available images from all over the web, including sites like DeviantArt, ArtStation, Getty Images, Pinterest, and many more.
Dozens of artists have spoken out about having their artwork taken without ever being asked or credited for it. Some alleged that AI is stealing from artists.
So what was the reaction of the developers? Prisma Labs CEO Andrey Usoltsev explained that Lensa AI art portraits “can never be described as exact replicas of any particular artwork.” According to him, the app generates images from scratch without “borrowing” from existing pieces of art.
Ricardo Baeza-Yates, director of research at the Institute for Experiential AI at Northeastern University, takes a different approach to soothe the waters. In his opinion, all depends on the number of the images used for creating AI art. If the AI-generated piece is based, for example, on 10 images, it’s likely to be considered plagiarism. However, if artificial intelligence uses 1,000 images to create a new piece, Baeza-Yates doesn’t see any cause for talking about art theft.
Mmh. Bold hypothesis. I could argue that stealing from 1,000 images is worse than stealing from ten, because you are not only harming ten artists but a hundred times as many… Doesn’t plagiarism remain plagiarism no matter how many you copy from?
Even if we accept that AI art isn’t plagiarism, we cannot deny that image generation systems only work because they have been trained with the artwork of real artists.
What are your thoughts on this? Is AI art theft or is it ok if you use so many pictures that you cannot make out the originals anymore? (Leading question, I admit...)
How does AI art affect artists?
Apart from using original artwork for fitting together new creations, AI art affects artists yet in other ways.
Artists have been raising ethical concerns because artificial intelligence also copies the styles of both alive and dead artists. The “new artwork” that AI generates can easily be mistaken for an original piece by the artist whose style it imitates. AI art generators ‘learn’ and use these particular styles without the artists even knowing or having given their consent.
The AI art generator trend made is super-easy to generate images in the style of the famous artists in history within seconds. Here is an example of Dublin at night in the style of Van Gogh’s ‘Starry Night’. (Image via DeviantArt)
Many artists who have spent years developing a unique art style feel that AI art is a threat to their livelihood. There is no way that human artists can possibly compete with AI systems to create high-quality work at such a fast pace. Many of us got into making art because we felt a calling to be artists. It is frustrating for artists to watch having their style copied without being able to do anything about it.
In my previous post on AI Art vs. Human Art, I already talked about the case of contemporary fantasy artist Greg Rutkowski. His style has become one of the most popular ones used for image generation in Stable Diffusion. While anyone can now create spectacular-looking imagery in Rutkowski’s signature style, the artist himself doesn’t receive any credit – let alone any compensation.
The ones who benefit are the developers behind the AI systems. They are the ones who get paid for the fakes of other artists’ original work. They charge subscription fees for those using their platforms.
Will AI take over artists?
The AI art generator trend has been swamping the web with images that are continuously becoming harder to distinguish from handmade work. Most people use artificial intelligence to create fan art of their favorite stars.
Many fear that this is the downfall of genuine creativity because they are anxious that AI will take over artists.
A lot of artists are worried about their creative future. They do not only see fan art and fake art styles as a problem, there are more areas where AI art could potentially replace artists. For example, digital art commissions. How many people are still going to pay for commissions if they can get similar results for much cheaper from an AI generator?
Are you worried about your creative future with all the machine-made imagery swallowing up the market?
AI might also take away jobs for artists when it comes to creating visuals for ads, magazines, events, marketing campaigns, book covers, websites, you name it. In the eyes of many artists, this kind of artificial technology deliberately bypasses human qualities like creativity, imagination, persistence, and artistic skill development. But it also willfully provides a way to avoid paying real artists.
Magazines like the Cosmopolitan already had AI-generated art on its cover, and websites like The Atlantic use AI-generated images at top of their articles, where they formerly used to have manmade art (and pay the artists for it). Unfortunately, many industries tend to value speed and cost-effectiveness over quality. A ‘just good enough’ machine-made image can satisfy the demands in a lot of cases.
Nevertheless, the technology is not quite there yet. Artificial images still have flaws that are almost immediately noticeable. For example, there are shadows where there shouldn’t be any, people have eyes that don’t match, extra fingers or toes, or an extra ear on their jacket. That means AI art – so far – still requires a human with digital art skills to fix things up.
Is AI artwork copyrighted?
The copyright to AI-generated artwork is not really the problem. Most image generation platforms agree that the ownership rights belong to the person who enters the prompts. Midjourney, for example, states in its Terms of Service that the user “owns all Assets you create with the Services.” Likewise, Stable Diffusion concedes all proprietary rights to the promptor.
The major concern about AI art is that it is presumably stealing from artists and using their works for new creations. Without any license, without any payment, and without any attribution to the artists whose work has been used.
How is that possible? AI-generated art is still a legal gray area. Unfortunately, the laws are not as hard and fast when it comes to the use of copyrighted artwork by artificial intelligence. AI art might fall under fair use laws, which makes it more complicated to enforce copyright claims.
The issue is similar with AI art replicating the style of particular artists. While individual artworks are subject to copyright, artistic styles are not. Already in 1972, US courts established that copyright law does not apply to an art style. On a similar note, the Australian Arts Law Centre believes that copying stylistic elements may not be considered copyright infringement.
However, Michael Bennett, an intellectual property lawyer, holds a different view. Bennett, currently working at the Institute for Experiential AI, sees good chances for artists like Greg Rutkowski to assert their ownership rights. After all, there is a strong resemblance between the respective AI-generated pieces and the artist’s original work.
Rutkowski keeps pushing to remove the artwork of living artists from the datasets used for training AI systems. His advocacy raised the attention of Daniela Braga, data scientist at the White House Task Force for AI Policy. Braga decided to bring the issue up to the White House office to find a legislative solution.
The crux of the matter is that there is neither an option for artists to opt out of having their work used for AI training purposes nor to “untrain” AI systems. Therefore, Braga considers “eradicating the whole model that was built around nonconsensual data usage.”
The last word in the AI art copyright debate is not said and we still need to see where it is going.
My guess? Mmh. Due to the copyright and ethical fustercluck, I could imagine that the names of living artists (maybe of dead ones as well) will eventually be banned from being used on AI systems. Just in the same way that the systems already ban the use of sensitive terms like “nude” or “blood.”
If that doesn’t happen, I think we will end up seeing artificial image generators as a toy that is there for people (especially non-artists) to play with and soon get bored of seeing yet another version of Spongebob in a Starry Night. ☺︎ But I speak more on that in my next post.
What’s your opinion? Do you think AI art is stealing from artists? Do you see it as a threat to your artistic career? Or are you totally chilled about it? ☺︎ Please leave comments and questions below or reach out to me directly. I’d love to hear from you! Make sure to sign up for my email list below so you don’t miss out on new blog posts and other cool stuff. ♥
For more about the topic, check out more posts from my AI Art vs Human Art series:
