The AI art debate is evolving at a lightning pace. In recent weeks, we've seen an AI-generated image win a fine art competition, we've seen a museum launch an exhibition of work created using DALL-E 2, and, in what appears to be another first, an artist has copyrighted work made using an AI art generator.
In what feels like an inevitable development in the debate around AI-generated art Kris Kashtanova has copyrighted a graphic novel that features images made using the platform Midjourney. It's ignited heated responses on social media, but is the news really as significant as it sounds? (If you need to catch up on how AI art generators work, see our piece on how to use DALL-E 2).
Kris Kashtanova's graphic novel Zarya of the Dawn is, well, novel. Firstly, it's a graphic novel made entirely of images that were generated using an AI art generator. We're not sure if that's a first, but it's at least one of the first. But what's caused the biggest stir is that Kashtanova has copyrighted the work with the Copyright Office of the USA, which appears to be a first.
Kashtanova wrote on Instagram that she took the decision to apply to the Copyright Office to "make a case that we do own copyright when we make something using AI." She stressed that she “was open about how it was made and put Midjourney on the cover page."
Kashtanova described her artwork as AI-assisted rather than created by AI. She wrote the story and designed the layout of the graphic novel, making the choices about how to put the images to together.
But copyrighting the work has proved to be particularly controversial because of her main character's apparent resemblance to the American actress Zendaya. Many AI art generators have been trained on images that include celebrity photographs, and AI artists sometimes use celebrity names in their prompts to ensure that characters in a series of images are consistent, so where does the owner of the images used to train the AI stand?
The answer to that question has proved murky enough to cause Getty Images took the decision to ban AI-generated images from its stock library. Other image libraries haven't – indeed, Kashtanova who describes herself as a Part-time Prompt Engineer as well as a designer and photographer shows on her Instagram that she has sold AI-generated imagery via Adobe Stock and Shutterstock.
so...AI thinks picturesque landscapes (technically, stolen art considering it analyzed images of orginal art from real artists and mimics them) + zendaya lookalike = comic book??? https://t.co/XSMU5LTd1BAugust 28, 2022
ice cold take: AI will never be able to make art. Unless AI becomes sentient, all AI art will just look like this, zendaya with some technically detailed backgrounds https://t.co/99qU8kksY9August 29, 2022
There should be legislation that makes it that AI art needs to be labelled as such and the training dataset has to be public. And that the art produced by the AI is owned by all copyright holders of the training data.September 30, 2022
Ideal AI art: “I want a pic of something, I’ll use an AI image generator that uses images where all parties involved consented to it’s use for commercial AI use, and I acknowledge that I did not create this nor own the copyright on the generated image”What’s happening: not thatSeptember 30, 2022
But some people have pointed out on Twitter (with their own mods of her comic) that Kashtanova's copyright registration might not be as meaningful as it seems. Registration is automatic and the question of whether it's valid would usually only come up if the registration is challenged legally. Secondly, her copyright is for "visual material" not "visual arts work". It seems the copyright is for the literary authorship of the graphic novel rather than the visual art contained therein.
It's similar to LHOOQ. If you add a moustache to a public domain work then only the mustache can be protected. Duchamp doesn't suddenly become the owner of copyright in the Mona Lisa.https://t.co/GU4Vgkk2KH.September 22, 2022
That aside, it's still theoretically possible to copyright AI art at the moment. The US Copyright Office has decided against allowing copyright to be given to an AI, with the AI registered as the author, but it hasn't ruled out granting copyright of AI-generated works to a human AI. This first would now seem to set a precedent that makes that possible, but that will only be tested if the registration is challenged. All the same, it's provoking some heated debate on Twitter, as shown in the tweets above.
Prepare to see a lot more of this kind of debate, especially now that Open AI has opened DALL-E 2 access to everyone. To learn more about what the technology can do, read about DALL-E 2 outpainting, which can expand pictures beyond their borders, and take a look at the best AI art generators compared.