41-year-old Jason Allen spent more than a hundred hours on Midjourney to create an award-winning image. Now he’s challenging the U.S. Copyright Office’s decision that he can’t copyright his work.
Two years ago, board game designer Jason Allen became fascinated with AI-generated images of surreal landscapes popping up on his Facebook feed and started experimenting with text-to-image AI programs himself. In May, he spent more than 100 hours instructing image generator Midjourney to create an elaborate illustration of women donning Victorian dresses and space helmets and attending a futuristic royal court. The image went on to win first prize in the Colorado State Fair for digitally manipulated photography a few months later.
But when he tried to copyright the image, the U.S. Copyright Office denied his application, saying that the work lacked “human authorship,” and that it wasn’t able to determine whether the prompts were “sufficiently creative.”
It was a slap in the face for Allen, who wrote 624 different text-based prompts to encourage Midjourney’s software to produce what he wanted, adjusting the style, composition, colors and tone of the image.
“It was hard to get the type of completions that I was looking for at the time using version two of Midjourney,” he told Forbes, adding: “There was a lot that went into it.”
Now, he is suing the agency and asking a federal court to reverse the Copyright Office’s decision.
The Copyright Office declined to comment on pending litigation. Midjourney did not respond to a comment request, but its website states that artists own all the images they create and can use them however they like. Its terms of service explicitly state that the company also perpetually and irrevocably owns the copyright licenses to reproduce, sublicense and distribute any works created through its software and has the right to distribute any content that is put into the system.
Allen told Forbes that after the agency denied copyright protection for the image, titled “Théâtre D’opéra Spatial” (French for “space opera theater”), his work was ripped off by other artists, and people have tried to sell the image as their own on platforms like Amazon, Etsy and NFT marketplace OpenSea. He believes it’s tied to the backlash from the artist community, which accused him of cheating in the Colorado State Fair competition in 2022 by using AI tools.
“There were several people who specifically said that they were setting out to steal my work and sell it as their own because there’s nothing I could do about it and they wanted to rub it in my face,” Allen said. One person, for instance, embedded the entire image into an AI-generated artwork of their own and wrote in a Facebook post that “They say that ‘great artists steal’, but as the IP never vested in Jason, no theft has occurred.”
Allen’s lawsuit comes at an important moment for artists and creators, amid a broad debate around the legality of companies scraping swathes of data from the internet to train generative AI software. In 2023, a group of artists filed a class action lawsuit against AI companies like Midjourney, as well as Stability AI and Runway, the creators of image generator Stable Diffusion, for allegedly using billions of copyrighted artworks without consent or compensation to train their AI models that now directly compete with the artists themselves. Last month, a court allowed the copyright infringement lawsuit to proceed, with a few claims dismissed, to discovery — a formal process where both parties exchange key information. For artists, this would allow them to get a peek at the models’ training data.
The companies have responded by asking the court to dismiss the claim entirely. Runway said artists have been unable to create exact copies of their works through Stable Diffusion or prove that it stores them. Stability AI has claimed that artistic styles can’t be copyrighted and the models themselves aren’t infringing on copyright because they are pieces of code rather than artworks. And, Midjourney’s lawyers have compared its AI models to a photocopier, making the argument that the underlying technology isn’t infringing copyright even if it does make exact replicas of another image.
Some of these questions have landed with the Copyright Office, which has been grappling with what AI means for ownership and intellectual property. The 150-year-old agency, housed within the Library of Congress, is trying to determine how centuries-old laws apply to contemporary problems that AI has created: examining whether the use of copyrighted works to train AI systems is a violation of law or if it falls under fair use, as many AI companies argue, and what types of works created with the help of AI can be copyrighted.
In March 2023, it issued a ruling that it won’t grant copyright registrations for works that are fully AI-generated. But that still leaves room for situations when works employ AI but also demonstrate a certain level of human authorship and creativity. In February, the agency said it had blocked works that it considered to be too much AI and not enough human input, while also registering more than 100 works that contain AI-generated materials like song lyrics, written text and visual elements.
Allen isn’t the only artist to contest the agency’s decision. In April, Elisa Shupe, an author who used ChatGPT to write a book that she self-published on Amazon, appealed and won a copyright for the selection and arrangement of AI-generated text, Wired reported. That means that the book as a whole cannot be copied legally, but individual sentences and sections of the book can be rearranged and published — similar to fully human-written books.
Allen said that even if the image he’d created hadn’t required hundreds of attempts and he’d only written one prompt for an image of a cat on a skateboard, for example, he believes the image should still be protected under copyright law.
“Rather than having to demonstrate the sweat of the brow and how much effort we put in, we don’t think that that’s a necessary factor in demonstrating human expression,” Allen said. “If you sat down with the intent to create a thing and using a machine as an assisting device, you created said thing. You are the author, you are the creator in our view.”
Allen now runs his own online gallery and sells fine art prints of illustrations he’s generated with the help of AI tools. He calls himself a digital creator rather than an artist and wants to be able to create hundreds of pictures with the help of artificial intelligence, all under the protection of copyright. “The thing about Midjourney is you don’t have to be an artist to create art anymore.”
