DALL-E Challenges Copyright Law With Controversial AI-Generated Art

"DALL E Flow" by AI Qu is licensed under CC BY-SA 4.0.

Artists are revered for their unique personal styles, but with the latest technological advances, everyone can be Picasso. OpenAI’s DALL-E grants users the ability to use an artificial intelligence system to generate art by simply inputting a word prompt. On Sept. 28, DALL-E went public, dropping the waitlist and giving everyone access to create AI-generated art within seconds. With one-and-a-half million users now generating two million images per day, it quickly became popularized with artists, designers and creators. This brings up the question: is AI-generated art considered copyright infringement? While it is easy to say that generations are blatant duplications of real art, the reality of the situation is that there are no definitive boundaries.

The AI is able to generate imitations of art styles. By prompting “creation of adam by van Gogh,” I was given the classic imagery of God’s touch giving Adam and the rest of humankind life, but it is generated with Vincent van Gogh’s iconic impressionist style. Added on with the sunflower in the corner that looks like it’s cut straight out of “Sunflowers,” it is undeniable that this AI has studied many of van Gogh’s paintings to be able to emulate his style. “Encouraging art students to mimic famous painting styles is a common way to let them learn the techniques and skills,” says Chad White, Bellevue College’s head of photography and art department chair. In a way, all art is inspired by other existing art. As long as it’s not an exact replica, there shouldn’t be an issue.

The real complication lies in the commercialization of AI-generated art. OpenAI’s policy states, “you may sell your rights to the Generations you create, incorporate them into works such as books, websites, and presentations, and otherwise commercialize them.” While their policy seemingly clarifies any legal trouble you might run into with selling the generation you created, this problem may be more perplexing. 

Selling your generations becomes more convoluted when you consider the copyrighted data in their databases. While OpenAI has filtered out images that were considered violent or sexual or promoted biases to train the AI on, there has been no mention on the copyrighted data that DALL-E might have been exposed to. Celebrities and political figures are banned from the AI generation due to them trying to mitigate the spread of misinformation, and when you try generating an image with a celebrity, it scrambles their features in an indistinguishable way. However, fictional characters were not given the same treatment. Prompting the AI with “oil painting of Darth Vader at Hogwarts” generated a Darth Vader look-alike. 

Difficulties may arise in trying to regulate and monitor where the data of your generations may have originated from. You might end up with a generation that had copyrighted material in it without even trying. The art the AI generates would never be fully in your control. The only solution users could use is to avoid prompts or uploads that might infringe on any copyrighted material. 

For the future of art, many would agree that AI may be part of the equation. White describes that currently artists are already trying to incorporate AI technology into their previously established art style. By feeding the AI art that could simulate their style, artists are using AI generation as a tool to create more.