In a relatively scathing opinion finding the plaintiffs’ Complaint “defective in numerous respects,” a district court judge has thrown out most of the claims a group of artists has asserted against AI platforms that allegedly used the artists’ copyrighted works without permission. The order in Andersen et al. v. Stability AI Ltd. provides an important preview on courts’ tolerance for AI-related copyright lawsuits moving forward—including a similar class action filed by actor/comedian Sarah Silverman and other authors.
As we previously wrote, the Andersen case relates to “Stable Diffusion,” an AI platform that generates images in response to user prompts. According to Plaintiffs, Stable Diffusion scraped the internet to copy and store billions of copyrighted images without consent or licenses to train the programs. (For another good summary of the case and the claims, check out this post from The Fashion Law).
Defendants filed separate motions to dismiss, and U.S. District Judge William Orrick stated during a hearing in July that he would likely grant the bulk of those motions, while leaving Plaintiffs the option to try to amend their complaint. On October 30, Judge Orrick issued a written decision to that effect, dismissing all but one claim for direct copyright infringement by Stability AI on behalf of just one plaintiff. Judge Orrick provided Plaintiffs with leave to re-plead their claims, however, stating that Plaintiffs need to allege more specifically how each defendant was involved in the claimed infringement. For example, discussing Midjourney, Judge Orrick wrote: “Plaintiffs need to clarify their theory against Midjourney — is it based on Midjourney’s use of Stable Diffusion, on Midjourney’s own independent use of training images to train the Midjourney product, or both?”
Importantly, Judge Orrick signaled that Plaintiffs are going to need to allege that the “Stable Diffusion” output images are “substantially similar” to Plaintiffs’ original artwork in order to prevail on certain of their copyright claims. As the judge put it, “I am not convinced that copyright claims based [on] a derivative theory can survive absent ‘substantial similarity’ type allegations.” He suggested that Plaintiffs’ might be able to allege additional facts to show “that Output Images can be so similar to plaintiff[s’] styles or artistic identities to be misconstrued as ‘fakes.’”
This decision provides some clarity to plaintiffs asserting AI-related copyright infringement claims, and will likely have an impact on other cases, including Silverman v. OpenAI. As we’ve previously written, the defendants in those cases raised the same “substantial similarity” question as in Andersen. Not surprisingly, the Silverman Defendants have already submitted the Andersen decision to the judge overseeing the Silverman case, claiming that the decision supports their motions to dismiss Silverman’s and her co-plaintiffs’ claims. We may see soon what impact, if any, the Andersen decision has in the Silverman cases: a motion hearing is currently noticed for December 7, 2023.
It will be interesting to see whether the court in Silverman adopts a similarly critical view of the plaintiffs’ claims in that case, as both Andersen and Silverman are serving as an important litmus test for the viability of copyright infringement claims against AI platforms. The cases continue to raise questions about whether the existing Copyright Act can support such claims, and—if it does not—whether the Act should be amended to address the plaintiffs’ alleged harms.