Let’s Change the Laws Governing AI Art
The laws governing AI art are outdated and need to be changed. The laws were written for a previous era, and don’t reflect the way individuals are creating and using AI art today. The laws governing AI writing also need to be revisited and revised, but I’ll address that in another article.
I first got involved in creating AI art about a month ago after reading an article about the Midjourney platform, and I thought I’d try it out on a children’s picture book, after working with illustrators for two dozen picture books for myself and clients. Initially, I created a story called Beyond the Beach about a boy who loses his family in a tsunami, goes to the beach to find them, and encounters a whale who takes him out to sea, where he meets a mermaid, who helps him get back. Using a series of prompts made up of short sentences describing each picture, I created 19 pictures for the book.
Then, since the images worked, I created images for three other books I already had written — Whoosh!, about a boy who goes on his first plane ride and sees all kinds of images through the plane window; A Dog’s Tale: Why Can’t I Be? about a Beagle who imagines being different types of dogs; and Zoo Do, about a boy who visits the zoo and imitates some animals he sees there. I picked these stories, because creating AI images works best when a story doesn’t follow a particular character or characters, since an AI prompt results in a variety of images and settings for a character, rather than creating a series of illustrations with the same character, as illustrators do.
Meanwhile, I began wondering about the big concern of many artists and writers: will AI replace them; will it cause them to lose jobs and clients, since AI can create images and copy so much more quickly and cheaply. For example, I created the images for each of my picture books in about 5 hours over two days for the cost of a $10 basic package on Midjourney, whereas hiring an illustration would cost about $500–700 and take the illustrator about two to four weeks to complete. Then, I created an article on this concern about replacement: Will AI Replace Writers and Artists? I even used AI generated text from WriterArc to write the whole article, except for my one-paragraph introduction, explaining how I had used AI to write the article. Finally, I used another AI program, CourseReel, to generate the text for a course on creativity consisting of 18 AI-written classes. Then, using AI, I turned that text into the script, narration, and images to produce a video for each class. Later I turned the slides in each video into a book with a similar name: How You Can Be More Creative, since I already had a book on creativity techniques: You Can Be More Creative.
Thus, while AI can replace writers and artists in some ways, it is also a tool that writers and artists can use to create other writing and art, just like any new technology has been adopted and used throughout history. So while AI writing and art may be disruptive in some ways, it is helpful in others and is here to stay.
Its staying power, in turn, raises two legal questions — how can AI art and writing be protected when it’s published from being copied by others, and how can AI be used legally, when it draws on the art and text already created by others. In other words, protecting one’s rights in any AI created art or writing raises copyright issues, while considering how AI can be used legally raises concerns about infringement. But now the laws affecting AI were developed before the AI has become widely popular today. Thus, these laws need to be changed, especially for artists, which I’ll discuss here.
Revisiting and revising these laws are especially critical now because of the large number of artists and others using AI programs to create art. When I first learned about AI art and created illustrations for my kids’ books, I thought I was doing something novel. I even submitted Whoosh! to publishers, announcing that I had uniquely created all the artwork using AI. Well, not only did I discover this wasn’t this unique, but one editor wrote that they wouldn’t publish anything with AI illustrations, because they couldn’t be copyrighted and potentially infringed on the work of other artists whose work had been incorporated into an AI program.
That’s when I discovered the huge number of AI artists. As of this writing, in the past few months, AI-based image generators, which enable anyone to create images by typing a few words or a sentence into a text, have attracted millions of users. For instance, Midjourney, which uses a Discord server, has over 3 million members, while DALL-E-2 has over 1.5 million users. And many other platforms, including DreamStudio, based on Stable Diffusion, and InteriorAI, are quickly signing up members, and Imagen from Google will soon be in the mix.
In turn, the AI programs draw their images from millions of images on the Internet. For example, DALL-E 2 has approximately 650 million images from the Internet paired with words of text, so when you enter a series of keywords or a sentence, AI will select corresponding images and combine them with other images or modify them in varying ways.
So now, given the rapid growth of AI and the multiple images it uses to generate art from the Internet, what is the legal status of using AI? Unfortunately, the laws are outdated, so they need to be changed to provide the necessary protections for copyright and to clarify what is infringement and what is not.
The problem with copyright law is that the U.S. Copyright Office has ruled that AI art can’t be copyrighted on the grounds that an image generated through AI lacks the “human authorship necessary to support a copyright claim. The Copyright Office ruled this way in a case brought by Stephen Thaler to copyright a series of images he created to simulate a near-death experience using his “Creativity Machine” algorithm. In both a 2019 decision and a February 2022 review by a three-person board, the Copyright Office rejected his copyright request based on it lacking human authorship.
Yet, while the AI images may be generated by a machine, there is definitely human authorship in creating the prompts that generate a series of images and in selecting the image or part of an image to be featured in a book, painting, or art piece. Even Thaler’s attorney, who plans to appeal the ruling, agrees. As he has told Art News, as cited in an article in Smithsonian Magazine by Jane Recker, “U.S. Copyright Office Rules A.I. Art Can’t Be Copyrighted, “We disagree with the Copyright Office’s decision and plan to appeal…AI is able to make functionally creative output in the absence of a traditional human author and protecting AI-generated works with copyright is vital to promoting the production of social valuable content. Providing this protection is required under current legal frameworks.”
Thus, whether AI art should be protected based on the input of a human to create it or on the output of a machine, the law should be changed to permit the copyright of AI-created works.
As for the infringement issue, there shouldn’t be any infringement if the data-mining process results in using an image or part of an AI-generated image which consists of multiple images or transforms the original image in a significant way, unless it is using a recognizable copyrighted, branded, or trademarked image, such as a Disney or Marvel character. Otherwise, the use of the image should be covered by the fair use exception that has long permitted the use of small sections of text or parts of images or transformed images.
Some artists would disagree, claiming that these AI images are created “on the backs of human artists, photographers, and designers, who are not seeing any benefit from these business models,” according to an August 11, 2022 article in Techol.lama, “Copyright Infringement in Artificial Intelligence Art” by Andres Guadamuz. Some artists have even called this use “theft” or “exploitation.”
But it is not. It is fair use or in many cases, the images are already available for use by anyone, such as the images that come from open access or public domain works, such as those with permissible licenses under Creative Commons. And still other works have been placed in the public domain or have fallen under copyright. But what of the other images which come from the Internet, which is the largest repository of images in the world? Should artists retain any rights in these images? Increasingly, according to Guadamuz, there is growing recognition that such data mining should be allowed under copyright “as fair use or fair dealing.”
An early example of a court case that reached this fair use result is the Google Books case, in which the Author’s Guild claimed that the Google Books Project infringed by creating the Google Books database through the mass digitization of millions of images. However, after over a decade of litigation, the right to use these images was affirmed when the Second Circuit Court ruled, and the Supreme Court let its ruling stand in 2016, that there was no copyright infringement. Rather, as the Court ruled, there was fair use because “The ultimate goal of copyright is to expand public knowledge and understanding.”
The one exception should be where AI generates a clearly recognizable branded or trademarked image, unless that image has been transformed in a way that is currently considered subject to copyright in art, such as when an artist creates a painting of a photographic image or creates variations on a comic book character. In some cases, there may be a copyright violation, in other cases not, depending on the work and what the artist has done to change it. Whatever restrictions already apply to such art created by artists should apply to AI created art. Otherwise, the fair use doctrine should apply, and this might be codified in the law so AI art, with these exceptions, is clearly protected and able to be copyrighted, once the copyright law is changed.
I’m hoping that this article might bring together a network of artists and writers who would like to see the law changed to provide copyright protection for AI created works and affirm that AI created works are covered by the fair use doctrine, apart from the exceptions noted for branded and trademarks art. Then, I hope this network can help to get lawmakers to support such legislation.
For more information, and to see links to the articles cited in this article, you can visit the Facebook group — The AI Revolution and Writers and Artists.
Gini Graham Scott, Ph.D. is the author of over 50 books with major publishers and has published 30 books through her company Changemakers Publishing and Writing (http://www.changemakerspublishingandwriting.com). She writes books and proposals for clients, and has written and produced over 50 short videos through Changemakers Productions. (http://www.changemakersproductions.com). Her latest book is I Was Scammed, available on Amazon at: https://www.amazon.com/Was-Scammed-Updated-Expanded-Becoming-ebook/dp/B09PNB38GJ. A paperback and hardcover are available, too. She is also the author of How to Find and Work with a Good Ghostwriter.