Some of the biggest AI news this week comes from a federal judge in Northern California, who denied critical parts of Stability AI, Midjourney, Inc., DeviantArt, Inc., and Runway AI’s bid to get a copyright-centric lawsuit waged against them by a group of artists dismissed. In an August 12 order, Judge William Orrick of the U.S. District Court for the Northern District of California dismissed Plaintiffs Sarah Andersen, Kelly McKernen, and Karla Ortiz’s claims of unjust enrichment, breach of contract, and violation of the Digital Millenium Copyright Act. But more important that the causes of action that the court dismissed (after previously dismissing some of the plaintiffs’ claims in October 2023), are the ones it kept in place, namely, the plaintiffs’ copyright infringement claims.
A bit of background: To date, the plaintiffs have set out a number of copyright infringement claims against the defendants – which are developers of platforms that use generative AI to transform users’ text prompts into images – based on at least two direct infringement theories …
> The “Model Theory,” which is based on the theory that Stability AI’s Stable Diffusion1.5 product – after it was trained – is “an infringing Statutory Copy” of the plaintiffs’ works or a “Statutory Derivative Work” because it represents a transformation of the plaintiffs’ works.
> The “Distribution Theory” of infringement, which is based on allegations that the defendants infringe the plaintiffs’ exclusive distribution rights because distributing Stable Diffusion 1.5 is equivalent to distributing the plaintiffs’ works.
The plaintiffs also make induced copyright infringement arguments, asserting that Stability AI is inducing infringement by distributing its generative AI models when any third-party downloads, uses, or deploys the models provided by Stability AI.
In refusing to decide the bulk of the plaintiffs’ case (i.e., their copyright infringement claims), the court found that Andersen, McKernen, and Ortiz sufficiently alleged that their works were included in the vast datasets used by the defendants to train the models that power Stable Diffusion and similar generative AI platforms.
With regard to the defendants’ attempts to sidestep Andersen, McKernen, and Ortiz’s copyright claims, Judge Orrick pointed out that Andersen and co. have identified their copyrighted works and described how these works were used in the process of training the defendants’ models. According to the court, the plaintiffs have made “plausible allegations showing why they believe their works” were included in the datasets at issue and used as training data for the defendants’ models, and also plausibly allege that the defendants’ platforms “produce images – when their own names are used as prompts – that are similar to plaintiffs’ artistic works.”
> “The plausible inferences at this juncture are that Stable Diffusion by operation by end users creates copyright infringement and was created to facilitate that infringement by design,” the court held.
In addition to refusing to decide on the plaintiffs’ Copyright Act claims, the court opted not to delve into the defendants’ fair use arguments. What the judge did say on this front is that “whether a fair use defense applies [is an] issue that must be tested on an evidentiary basis.”
Looking beyond copyright claims, what still in play following the latest round are false endorsement and trade dress infringement claims, which the plaintiffs have asserted “in place of the right of publicity claims asserted against Midjourney that were dismissed in [the court’s] October 2023 order.”
> False Endorsement: In furtherance of their false endorsement claim, some of the plaintiffs allege that their names appeared on the list of 4700 artists posted by Midjourney’s CEO on Discord, the platform where Midjourney’s AI product operates. Midjourney’s CEO promoted the list as describing the various styles of artistic works that the AI product could produce. The plaintiffs also allege that Midjourney has published user-created images that incorporate their names on Midjourney’s “showcase” site.
> Trade dress infringement: According to the plaintiffs, Midjourney used their names in connection with its use of a model that has been trained on their work and allows users to create works capturing the “trade dress of each of the plaintiffs that is inherently distinctive in look and feel as used in connection with their artwork and art products.”
Letting both claims move ahead, the court held, among other things, that: (1) the plaintiffs have plausibly alleged that the use of their names by Midjourney in the List and showcase misleads consumers regarding source and endorsement, and whether or not a reasonably prudent consumer would be confused or misled by the names list and showcase to conclude that the included artists were endorsing the Midjourney product can be tested at summary judgment; and (2) the combination of identified elements and images, when considered with plaintiffs’ allegations regarding how the Midjourney model works as a trade dress database, and Midjourney’s use of the plaintiffs’ names in its Midjourney name list and showcase, provide sufficient description and plausibility for plaintiffs’ trade dress claim.
(For a dive into trademark causes of action in generative AI cases, which are largely dominated by copyright and Digital Millennium Copyright Act claims, you can find that right here.)
THE BIGGER PICTURE: The case at hand is part of a larger pool of litigation that centers on the rise and increasing adoption of generative AI. The outcomes in these cases, many of which are still in relatively early stages, stand to have a significant impact on the workings of companies like Stability AI, OpenAI, etc., which have been widely accused of building very valuable generative AI platforms by making unauthorized and allegedly infringing use of sizable amounts of copyright-protected materials – from imagery to books and newspaper articles. As such, how exactly companies in the generative AI space can function (and how they can continue to build and train the models necessary to power their AI platforms) stands in the balance.
The case is Sarah Andersen, et al., v. Stability AI LTD., et al., 3:23-cv-00201 (N.D. Cal.).