.jpg)
Work by Major Artists, Animators, Cartoonists Allegedly Trained AI
By Camilla Alvarez Chow
Screenshot of JON LAM‘s post on X (formerly Twitter) exposing Midjourney software developers’ conversation.
San Francisco-based research lab Midjourney is once again in hot water. While many were ringing in the new year, a list containing the names of 16,000 artists whose work Midjourney allegedly used to train its generative, text-to-image artificial intelligence program, was circulated on X (formerly Twitter).
The hefty 24-page document, originally produced on Google Sheets, includes modern and contemporary artists such as Japanese avant-gardist Yayoi Kusama, British-Indian sculptor Anish Kapoor, late Korean-American video artist Nam June Paik, late Turkish painter and sculptor Selma Gürbüz, and Egyptian muralist Alaa Awad. Artists in the entertainment field were also implicated, such as Studio Ghibli founder Hayao Miyazaki, Indonesian freelance illustrator and game concept artist Fajareka Setiawan, late Indian political cartoonist Ajit Ninan, and Korean manhwa artist and writer Ha Il-kwon.
Jon Lam, a senior storyboard artist at Riot Games, was one of the artists to chime in, tweeting several screenshots of Midjourney software developers on Discord discussing the creation of a database of artists, time periods, styles, genres, movements, mediums, and techniques to train its AI model. The range of visual creators includes major 19th-century figures like the post-Impressionists Paul Cézanne, Paul Signac, and Vincent van Gogh; iconic 20th-century artists from Pablo Picasso to Andy Warhol and Cy Twombly; and Americana illustrators such as Norman Rockwell and Walt Disney.
The list was also included as part of a recent amendment to a class-action complaint by American artists Sarah Andersen, Kelly McKernan and Karla Ortiz against Stability AI, Midjourney, and DeviantArt, originally filed in January 2023. After a judge dismissed the suit, the artists submitted 455-pages worth of supplementary evidence in late November, including the now-published list. The Google file has since been restricted, but a duplicate version has been uploaded to the Internet Archive.
Concerns over AI firms “stealing” artworks to train image generators has encouraged University of Chicago researchers to create “Nightshade,” a web-based tool for artists that “poisons” image sets, confusing text-to-image outputs. In addition, artists have been encouraging their peers to seek legal representation and register their work with copyright offices—a lack of which contributed to Anderson, McKernan, and Ortiz’s suit originally being dismissed. As expected, the performing arts are safe from being scraped due to their experiential nature, however the same cannot be said of commercial artists in the comic book, animation, and games industry.
The contentious battle against AI training on copyrighted material has governments scrambling over how to regulate this emerging technology. Japan drafted guidelines this past November to ensure transparency and necessary protections. As livelihoods are threatened and years of training become reduced to ones and zeroes, however, the release of this list suggests that artist solidarity is needed now more than ever.
Camilla Alvarez Chow is an editorial assistant at ArtAsiaPacific.
