argument: Commenti/Comments - Intellectual Property Law

In the case of Sarah Andersen et al. v. Stability AI Ltd. et al., the United States District Court for the Northern District of California grapples with the intersection of copyright law and artificial intelligence, focusing on allegations that AI software Stable Diffusion used copyrighted artworks without permission to create new images. The defendants' motion to dismiss highlighted challenges in establishing direct copyright infringement, given the transformative nature of AI-generated art. The court's order permits plaintiffs to amend their complaint, addressing deficiencies in demonstrating copyright infringement and involvement of each defendant. This ruling underscores the complexity of applying traditional copyright principles to AI outputs and emphasizes the necessity for precise legal articulation in cases involving advanced technologies. The evolving nature of the right of publicity in the AI context suggests a need for legal frameworks to adapt to the realities of digital and AI-driven creations, marking a significant moment in the legal discourse surrounding copyright law and artificial intelligence.

» view: il documento ( Sarah Andersen et al. v. Stability AI Ltd. et al. case (23-cv-00201-WHO))

written by Marco Perilli

In the intricate labyrinth of legal issues surrounding artificial intelligence and copyright law, we believe that one of the major focal point of debate is the training of AI systems for the creation of artistic works using copyrighted materials and the consequential transformative process that renders original works unrecognizable in AI-generated outputs. The Sarah Andersen et al. v. Stability AI Ltd. et al. case (23-cv-00201-WHO) before the United States District Court, Northern District of California, has become a touchstone in this ongoing discourse.

In the landmark of such case, the United States District Court for the Northern District of California was confronted with a complex legal challenge that sits at the intersection of copyright law and artificial intelligence technology. The plaintiffs, a group of artists including Sarah Andersen, Kelly McKernan, and Karla Ortiz, initiated a class action lawsuit against Stability AI Ltd., DeviantArt, Inc., and Midjourney, Inc., alleging that their copyrighted artworks were used without authorization to train an AI software product known as Stable Diffusion. This AI system subsequently generated new images, allegedly "in the style" of specific artists, which the plaintiffs argued constituted copyright infringement, among other legal violations.

The defendants responded with several motions, including motions to dismiss the complaint for failing to state a claim upon which relief could be granted (essentially, this means the defendants believed the plaintiffs' complaint did not meet the necessary legal standards to proceed with a case), and a special motion to strike under California's anti-SLAPP statute, aimed at preventing litigation that might impair valid exercises of the freedom of speech.  SLAPP (Strategic Lawsuit Against Public Participation) suits are frivolous lawsuits intended to silence free speech through legal harassment. Anti-SLAPP laws were enacted to counter SLAPPs by allowing quick dismissal of meritless suits that threaten First Amendment rights. California's anti-SLAPP law lets defendants file a special motion to strike and dismiss sham SLAPP cases. If granted, the suit is dismissed and attorney's fees can be recovered. Anti-SLAPP laws protect free speech by stopping abusive lawsuits that seek to stifle criticism.

The Court's resulting order on these motions is a meticulous dissection of the plaintiffs' allegations and the legal standards that govern them. The order grants the defendants' motions to dismiss in most respects, but it also provides the plaintiffs with the opportunity to amend their complaint to address the deficiencies identified by the Court.

The Court identified multiple areas where the complaint was found lacking. One of the primary issues was the plaintiffs' failure to provide clear articulation of the legal assumptions underneath their claims against each defendant. Copyright infringement requires a demonstration that the defendant copied protected elements of the plaintiff's work, which was not clearly established in the initial complaint. Additionally, the plaintiffs needed to delineate how each defendant was involved in the alleged infringement — a task complicated by the nature of AI and the distributed responsibility in the development and deployment of Stable Diffusion’s AI Systems.

The plaintiffs alleged that their copyrighted images were among the billions of images used to train Stable Diffusion. This training process, they argued, created derivative works of their original art. The Court, however, noted that the plaintiffs admitted that the output images from Stable Diffusion were unlikely to closely match any specific images from the training data. This admission, the Court suggested, undermined the plaintiffs' claims of direct copyright infringement, as it points to a lack of substantial similarity — a key element of copyright infringement claims.  We would like to emphasize the significance of this aspect. Proving that a copyrighted work forms the basis for a creation by generative AI systems is akin to what we call in other areas of law "probatio diabolica" or "diabolical proof." This is essentially a near-impossible burden of proof because the transformation often renders the original work unrecognizable. Otherwise, it would be considered plagiarism, a legal concept that allows for somewhat broader margins of interpretive leeway but is still more readily identifiable. Additionally, it's widely acknowledged that the internal processing mechanisms of AI systems are often inscrutable, even to the operators of these systems themselves.

The Court also addressed the plaintiffs' claims regarding the removal of, or alteration to, their copyright management information (CMI). The plaintiffs argued that the defendants violated their rights by failing to maintain this information in the AI-generated images. The Court called for a more detailed explanation of how the defendants' actions led to the removal or alteration of CMI, which is decisive for establishing a claim under the Copyright Act's provisions against the falsification or removal of copyright information. In this case, the artists' arguments appear to be not sufficiently solid from a procedural standpoint. Ma

Regarding the right of publicity claims, the plaintiffs contended that the defendants' use of their images without consent amounted to a misappropriation of their likeness. The "right of publicity" in this context refers to the legal right of an individual to control and profit from the commercial use of their name, image, likeness, or other aspects of their identity. The Court found that the plaintiffs did not sufficiently explain how their likenesses were used for commercial purposes — an essential component of a right of publicity claim. Again, the order highlighted the need for a more concrete factual basis to support such a claim.

The Court deferred ruling on DeviantArt's special motion to strike under California's anti-SLAPP statute, which protects defendants from lawsuits that aim to silence or intimidate them on issues of public interest. The Court chose to wait for the amended complaint before determining whether the anti-SLAPP statute applied.

_____________________

The order is noteworthy for several reasons. First, it emphasizes the need for plaintiffs to clearly articulate their claims when dealing with emerging technologies like AI. The complexity of AI systems and their multifaceted interactions with copyrighted materials demand a high level of precision in legal arguments.

Second, the order highlights the challenges inherent in applying traditional copyright concepts to the outputs of AI systems. The distinction between direct and derivative infringement becomes blurred when an AI system can generate new works that are reminiscent of, but not substantially similar to, the training data. In U.S. copyright law, specifically within the California jurisdiction, "direct infringement" occurs when someone directly copies or uses a copyrighted work without permission. This is a straightforward violation where the infringing work is an exact or substantially similar copy of the original. "Derivative infringement," on the other hand, involves creating a new work that is based on or derived from the copyrighted work, without authorization. This requires transforming the original into a new expression, story, or piece that still retains recognizable elements of the copyrighted material.

Third, the order discusses the evolving nature of the right of publicity in the context of AI. As AI systems increasingly generate realistic images of people, the legal framework around the commercial use of an individual's likeness may need reconsideration. Additionally, at least for European Union, where a most advanced personal data protection regulation is in force, this scenario could open new frontiers in analysis and protection from a personal data protection viewpoint, highlighting the need for legal adaptations to address privacy concerns alongside copyright and publicity rights.

Finally, the order sets the stage for further legal discourse on the balance between protecting the rights of copyright holders and fostering innovation in AI. The Court's decision to allow the plaintiffs to amend their complaint to address the specific issues outlined: a clearer demonstration of copyright infringement, more precise articulation of the role each defendant played in the alleged violations, detailed claims regarding the removal or alteration of CMI, and stronger factual support for the right of publicity claims.

The court found the plaintiffs' complaint to be defective in several ways and largely granted the defendants' motions to dismiss. However, the plaintiffs were given leave to amend their complaint to provide clearer theories of how each defendant separately violated their copyrights, removed or altered their copyright management information, or violated their rights of publicity, with plausible supporting facts.

The Andersen case serves as a pilot case for similar disputes likely to arise as AI becomes more prominent in the creation of artistic works. The outcome of this case and the legal principles it establishes will have far-reaching implications for artists, technology companies, and the boundaries of copyright law in the digital age.

February 8 Update:

On November 29, 2023, Plaintiffs filed an amended complaint omitting the previously asserted right of publicity claims.

Subsequently, Defendant DeviantArt renewed its special motion to strike the now-omitted right of publicity claims, and further sought recovery of attorneys' fees.

Thereafter, on February 8, 2024, the District Court denied Defendant DeviantArt's renewed anti-SLAPP motion on the grounds that Plaintiffs' action qualified for the public interest exemption under Cal. Civ. Proc. Code § 425.17(b). In the same order, the Court denied DeviantArt's request for attorneys' fees.