A federal judge in San Francisco found Monday (Oct. 30) that Stability AI could not dismiss a lawsuit alleging that it had “trained” its platform on copyrighted photographs, though he did side with AI businesses on critical issues.
Judge William Orrick found many flaws in the lawsuit’s assertions in an early-stage decision in a carefully watched case, and he dismissed some of the case’s claims. However, he permitted the case to proceed on its main allegation: that Stability AI constructed its tools by stealing a large quantity of copyrighted works.
“Plaintiffs have adequately alleged direct infringement based on the allegations that Stability downloaded or otherwise acquired copies of billions of copyrighted images without permission to create Stable Diffusion, and used those images to train Stable Diffusion,” the judge stated in his ruling.
The decision came in one of numerous cases brought against AI companies for using copyrighted content to train their models. Authors, comedians, and visual artists have all brought lawsuits against businesses such as Microsoft, Meta, and OpenAI, arguing that such unlawful use by the fast-growing industry violates copyright law.
The first such complaint involving music was brought last week by Universal Music Group and others, who claimed that Anthropic PBC was infringing copyrights en masse by using “vast amounts” of music to teach its software how to spew forth new lyrics.
Rulings in previous AI copyright cases could provide vital direction on how courts will handle such legal concerns in the future, perhaps influencing how UMG’s lawsuit and others like it play out.
The verdict was made in a class action lawsuit brought by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz against Stability AI Ltd. for their Stable Diffusion — an AI-powered image generator. Midjourney Inc. and DeviantArt Inc. were both named in the complaint as companies that employ Stable Diffusion as the foundation for their own image generators.
Judge Orrick dismissed many of the lawsuit’s claims in his decision. He dismissed McKernan and Ortiz from the case and directed the plaintiffs to re-file an amended version of their case that included substantially more detail concerning the particular charges against Midjourney and DeviantArt.
The judge also questioned whether every “output” image created by Stable Diffusion would be a copyright-infringing “derivative” of the photos used to train the model – a decision that may significantly limit the scope of the claim. The judge suggested that such images would only be infringing if they were “substantially similar” to a specific training image.
However, Judge Orrick did not include any such criticisms for the core complaint that Stability AI violated Andersen’s copyrights by using them for training without authorization – the main allegation at the heart of all AI copyright lawsuits, including the one filed by UMG. In future litigation, Andersen will need to prove that such an accusation is accurate, but the judge ruled she should be given the opportunity.
“Even Stability recognizes that determination of the truth of these allegations — whether copying in violation of the Copyright Act occurred in the context of training Stable Diffusion or occurs when Stable Diffusion is run — cannot be resolved at this juncture,” Orrick stated in the final paragraph of his ruling.
Stability AI, Midjourney, and DeviantArt attorneys did not respond to demands for comment. The artists’ attorneys commended the judge for allowing their “core claim” to proceed and onto “a path to trial.”
“As is common in a complex case, Judge Orrick granted the plaintiffs permission to amend most of their other claims,” plaintiffs’ attorneys Joseph Saveri and Matthew Butterick said following the verdict. “We’re confident that we can address the court’s concerns.”
Stay Updated about the latest technological developments and reviews by following TechTalk, and connect with us on Twitter, Facebook, Google News, and Instagram. For our newest video content, subscribe to our YouTube channel.
