Viral Lists Reveal Artists Whose Work May Have Trained an A.I. Art Generator

Artists are accusing the developers of an artificial intelligence art generator of ripping off their work after a list containing the names of thousands of creatives recently began circulating online.

They claim the list confirms what they had long suspected: that Midjourney—the company behind a popular text-to-image generative A.I. technology—used their artwork without their permission to train its art generator.

The list in question contains the names of 4,700 individuals. The artist Jon Lam shared it online in late December, but it was initially included in a court exhibit in November as part of a lawsuit against Midjourney and several other generative A.I. companies.

A larger spreadsheet titled “Midjourney Style List” is also circulating online. The spreadsheet reportedly contains the names of some 16,000 “proposed additions.” (Though its owners have made it inaccessible, a version of the sheet still exists on the Internet Archive’s Wayback Machine.)

David Holz, Midjourney’s founder, purportedly shared the list on Discord in 2022, reports NBC News’ Angela Yang. 

The lists feature painters, cartoonists, filmmakers, animators, sculptors and many other types of creatives, both living and dead. They include famous figures such as Tim Burton, Pablo Picasso, Norman Rockwell, Walt Disney and Banksy.

Many lesser-known artists also made it onto the lists, including a 6-year-old, reports ARTnews’ Karen K. Ho.

Together, the lists have reinvigorated the debate around whether A.I. should be allowed to imitate the work of real-life artists—and, perhaps more importantly, make money off those imitations. Legal and ethical questions about copyright infringement and consent continue to swirl, with artists calling for greater regulation of A.I. art generators and related technologies. Some creatives are asking lawmakers—and the courts—to help settle the issue.

Tools like Midjourney use machine learning models to study vast amounts of data, including online images of artwork and stills from movies and video games. Then, when a user inputs a prompt, like “Paint the Faroe Islands in the style of Vincent van Gogh,” the A.I. uses its training to produce an image that matches the request. Artists argue that the A.I. platforms are spitting out replicas of their existing work—much of which is copyrighted—rather than using it as inspiration to create something new.

In a 2022 interview with Forbes’ Rob Salkowitz, Holz acknowledged that the company did not seek consent from living artists to use their work while training Midjourney’s generative A.I. technology. The company also did not get permission to use copyrighted work, he told the publication.

“There’s no way to find a picture on the internet and then automatically trace it to an owner and then have any way of doing anything to authenticate it,” Holz said.

When asked if artists could opt out of having their work included in the training data, Holz responded that the company was “looking at that.”

Some artists are taking matters into their own hands. For example, researchers at the University of Chicago have created programs that make images difficult for A.I. tools to scrape.

In the meantime, as the legal landscape evolves, “documents like the ‘Midjourney Style List’ [can help] shed light on the actual processes of converting copyrighted artwork into A.I. reference material,” writes the Art Newspaper’s Theo Belci.

Sarah Kuta | READ MORE

Sarah Kuta is a writer and editor based in Longmont, Colorado. She covers history, science, travel, food and beverage, sustainability, economics and other topics.