Leaked Midjourney artist database could be a moment of reckoning for AI art

Ever since generative AI models entered the mainstream in 2022, controversy has plagued the tech. Questions surrounding ethics and copyright have only increased in ferocity in recent months, with the likes of even Adobe accused of copyright infringement by illustrators. And it's about to get a lot worse.

A newly leaked database titled 'Midjourney Style List' has been found to contain the names of over 16,000 artists whose work the generative model is believed to have been trained on. Spanning 24 pages, the document names Banksy, David Hockney, Walt Disney and many more, and is already provoking outrage online. 

While the original copy of the Google document is no longer accessible, it is viewable via the Internet Archive's Wayback Machine, and is included as part of a lawsuit against Stability AI, Midjourney, and DeviantArt. It was originally shared by Midjourney David Holz in 2022. 

As ARTnews has spotted, the list of names even includes a six year-old child. And there are countless artists on Twitter (sorry, X) noting that their own names is on there. And, of course, there are the names of illustrators and designers at major brands including Hasbro and Nintendo. 

@controllaxyz

♬ original sound - Controlla

"This list includes over 16,000 artists, many of whom barely make a living from their work," TikTok user Controlla says in a recent video (above). "I'm honestly disgusted that I've used this product in the past. This is another blatant theft happening in the AI industry."

AI image generation has come a terrifyingly long way in just one year, and it's becoming increasingly difficult to tell when an image has been artificially generated. The leaked document does seem to add credence to the fear of artists who believe their styles have been mimicked. "I'm against the use of my art in the databases used for generative AI," illustrator Loish recently told Creative Bloq. "I think so-called ethical AI stock images should not mimic the styles of artists who did not opt in to the use of their art in such a way. It's copyright infringement, in my eyes."

Indeed, the lawsuit could prove a moment of reckoning for Midjourney, and for the AI industry in general. While the likes of Adobe Firefly are allegedly only trained on commercially available images, this is the first time a model appears to have been (inadvertently) exposed as having been trained on countless artists' styles.