Artists have secured a small but meaningful win in their lawsuit against generative artificial intelligence art generators in what’s considered the leading case over the uncompensated and unauthorized use of billions of images downloaded from the internet to train AI systems. A federal judge refused to acknowledge that the companies can avail themselves of free speech protections and stated that the case is in the public interest.

U.S. District Judge William Orrick, in an order issued last week, rebuffed arguments from StabilityAI and Midjourney that they are entitled to a First Amendment defense arising under a California statute allowing for the early dismissal of claims intended to chill free speech. They had argued that the suit targets their “speech” because the creation of art reflecting new ideas and concepts — like those conveyed in text prompts to elicit hyper-realistic graphics — is constitutionally protected activity.

Related Articles

Metalwork Colossus by Hyan Tran, age 6, for a special drop of Magic the Gathering benefitting the Seattle Children's Hospital in 2021.

Database of 16,000 Artists Used to Train Midjourney AI, Including 6-Year-Old Child, Garners Criticism

Judge Dismisses Several Copyright Allegations Against AI Companies in Artist Class Action

The suit, filed last year in California federal court, targets Stability’s Stable Diffusion, which is incorporated into the company’s AI image generator DreamStudio and allegedly powers DeviantArt’s DreamUp and Midjourney.

In October, the court largely granted AI art generators’ move to dismiss the suit while allowing some key claims to move forward. It declined to advance copyright infringement, right of publicity, unfair competition and breach of contract claims against DeviantArt and Midjourney, concluding the allegations are “defective in numerous respects.” Though a claim for right of publicity was not reasserted when the suit was refiled, DeviantArt moved for its motion to strike the claim for good to be decided so it could recover attorney fees and resolve the issue, which could impact other cases in which AI companies assert First Amendment protections. Artists, in response, cried foul play. They stressed that the companies are abusing California’s anti-SLAPP law and attempting to “strongarm and intimidate [them] into submission.”

The right of publicity claim concerned whether AI art generators can use artists’ names or styles to promote their products. The suit argued that allowing the companies to continue doing so cuts into the market for their original works.

Orrick sided with artists on the issue of whether the companies can dismiss the claim under the state’s anti-SLAPP statute, finding that the “public interest exemption is met here.” He noted that the claim was initially dismissed because the suit failed to substantiate allegations that the companies used the names of Sarah Andersen, Kelly McKernan or Karla Ortiz — the artists who brought the complaint — to advertise their products.

“Had plaintiffs been able to allege those facts, they would have stated their claims,” the order stated. “That does not undermine that their original right of publicity claims were based on the use of their names in connection with the sale or promotion of DreamUp, a type of claim that would undoubtedly enforce California’s public policy to protect against misappropriation of names and likenesses.”

Lawyers for the artists have reserved the right to reassert their right of publicity claim pending discovery.

Though the court in October dismissed most of the suit, a claim for direct infringement against Stability AI was allowed to proceed based on allegations that the company used copyrighted images without permission to create its AI model Stable Diffusion. One of its main defenses revolves around arguments that training its chatbot does not include wholesale copying of works but rather involves developing parameters — like lines, colors, shades and other attributes associated with subjects and concepts — from those works that collectively define what things look like. The issue, which may decide the case, remains contested.

In December, scrutiny around Midjourney intensified after its latest update no longer remixed images and figures enough to obfuscate what legal experts said is clear copyright infringement. Users could prompt the AI art generator to return nearly exact replicas of frames from well-recognized films. One example: When prompted with “Thanos Infinity War,” Midjourney responded with an image of the purple-skinned villain in a frame that appears to be taken straight from the movie or promotional materials. The chatbot can also seemingly replicate most animation styles, generating picture-perfect characters from an array of titles, including DreamWorks’ Shrek, Pixar’s Ratatouille and Warner Bros.’ The Lego Movie.