Artists Face Setback in AI Copyright Infringement Case

A federal judge largely dismissed artists' claims against AI art generators over alleged copyright infringement, with one key claim against Stability AI set to proceed.

In a legal battle concerning artificial intelligence (AI) and intellectual property, artists have recently encountered a setback in their lawsuit against AI art generators over allegations of copyright infringement.

The case involves several artists who have accused Stability AI, DeviantArt, and Midjourney of using billions of unauthorized internet images, potentially including their copyrighted works, to train AI systems. The artists argue that these AI models, specifically Stability’s Stable Diffusion, unlawfully repurpose their intellectual property.

On Monday, U.S. District Judge William Orrick dismissed most claims against Midjourney and DeviantArt, suggesting that the allegations were “defective in numerous respects.” In his dismissal of the claims, Judge Orrick considered the plaintiffs' theory on how training images are stored in Stable Diffusion and utilized by DeviantArt and Midjourney to be “unclear,” – whether the AI systems in question directly contain copies of the copyrighted images or if the artists can demonstrate infringement without producing identical material generated by the AI tools. He suggested that the plaintiffs clarify their stance on the matter and provide facts to support their allegations.

The essence of the lawsuit is whether the artists can prove that their works were used to train Stability's AI system, which task is particularly challenging as the training datasets used by AI models are often opaque, making it almost impossible to establish clear evidence of infringement.

The artists also face the challenge of demonstrating that AI-generated works are identical or “substantially similar” to their copyrighted material, a requirement for a successful copyright infringement claim. The judge expressed skepticism regarding the viability of copyright claims based on a derivative theory without substantial similarity allegations.

However, one claim for direct infringement against Stability AI has been permitted to move forward. The allegation is against Stability AI that was accused of using copyrighted images without consent to create Stable Diffusion, a component of its AI image generator, DreamStudio. Stability denies these claims, asserting that training its model does not involve copying but merely developing parameters — such as lines, colors, and shades — from the images to recognize and generate visual concepts. Although some artists withdrew their claims due to unregistered copyrights, the lawsuit will proceed with the claims related to artist Sarah Anderson's registered works. Anderson used the website haveibeentrained.com to substantiate her allegations.

The lawsuit has attracted attention amidst broader discussions on AI ethics and intellectual property rights. The case remains closely watched by the creative and tech communities, given its potential to set precedents for future interactions between AI and intellectual property law. While this initial ruling is a setback for the artists, the judge has allowed room for the plaintiffs to amend and clarify their claims, indicating that this legal battle is far from over.


Previous
Previous

The MPA Warns of “Inflexible Rules” around AI and Copyright

Next
Next

The WGA East Members Petitions against AI Takeover in Journalism