From Casetext: Smarter Legal Research

Andersen v. Stability AI Ltd.

United States District Court, Northern District of California
Oct 30, 2023
23-cv-00201-WHO (N.D. Cal. Oct. 30, 2023)

Summary

granting motion to dismiss with leave to amend

Summary of this case from Lawrence v. Finicity Corp.

Opinion

23-cv-00201-WHO

10-30-2023

SARAH ANDERSEN, et al., Plaintiffs, v. STABILITY AI LTD., et al., Defendants.


ORDER ON MOTIONS TO DISMISS AND STRIKE

RE: DKT. NOS. 49, 50, 51, 52, 58

WILLIAM H. ORRICK UNITED STATES DISTRICT JUDGE

Artists Sarah Anderson, Kelly McKernan, and Karla Ortiz filed this putative class action on behalf of themselves and other artists to challenge the defendants' creation or use of Stable Diffusion, an artificial intelligence (“AI”) software product. Plaintiffs allege that Stable Diffusion was “trained” on plaintiffs' works of art to be able to produce Output Images “in the style” of particular artists. See generally Compl., Dkt. No.1. The three sets of defendants ((i) Stability AI Ltd. and Stability AI, Inc. (“Stability”); (ii) DeviantArt, Inc.; and (iii) Midjourney, Inc.) have each filed separate motions to dismiss and DeviantArt has filed a special motion to strike under California Code of Civil Procedure section 425.16. Finding that the Complaint is defective in numerous respects, I largely GRANT defendants' motions to dismiss and defer the special motion to strike. Plaintiffs are given leave to amend to provide clarity regarding their theories of how each defendant separately violated their copyrights, removed or altered their copyright management information, or violated their rights of publicity and plausible facts in support.

BACKGROUND

Plaintiffs allege that Stability created and released in August 2022 a “general-purpose” software program called Stable Diffusion under a “permission open-source license.” Compl. ¶¶ 1, 33, 52, 53. Stability is alleged to have “downloaded of otherwise acquired copies of billions of copyrighted images without permission to create Stable Diffusion,” known as “training images,” Id. ¶¶ 2, 3, 25-26. Over five billion images were scraped (and thereby copied) from the internet for training purposes for Stable Diffusion through the services of an organization (LAION, Large-Scale Artificial Intelligence Open Network) paid by Stability. Id. ¶¶ 57, 101, 104; see also id. ¶¶ 2, 25 (defining “Training Images”). Stability's founder and CEO “publicly acknowledged the importance of using licensed training images, saying that future versions of Stable Diffusion would be based on ‘fully licensed' training images. But for the current version, he took no steps to obtain or negotiate suitable licenses.” Id. ¶ 106.

Stable Diffusion is alleged to be a “software library” providing “image-generating services” to products produced and maintained by the defendants including “DreamStudio, DreamUp, and on information and belief, the Midjourney Product.” Id. ¶¶ 23, 50, 65. Consumers use these products by entering text prompts into the programs to create images “in the style” of artists. The new images are created “through a mathematical process” that are based entirely on the training images and are “derivative” of the training images. Id. ¶¶ 4, 5, 24. Plaintiffs admit that “[i]n general, none of the Stable Diffusion output images provided in response to a particular Text Prompt is likely to be a close match for any specific image in training data. This stands to reason: the use of conditioning data to interpolate multiple latent images means that the resulting hybrid image will not look exactly like any of the Training Images that have been copied into these latent images.” Id. ¶ 93. Plaintiffs also allege that “[e]very output image from the system is derived exclusive from the latent images, which are copies of copyrighted images. For these reasons, every hybrid image is necessarily a derivative work.” Id. ¶ 95.

DreamStudio is Stability's product, also released in August 2002; it functions as an “user interface” accessing “a trained version of Stable Diffusion.” Id. ¶¶ 33, 55. Use of DreamStudio is billed in packages of credits that can be used to create images. Id. ¶ 56.

Defendant DeviantArt was founded in 2000 and has primarily been known as an “online community” where digital artists post and share their work. Id. ¶¶ 35, 62. 114. Deviant Art released its “DreamUp” product in November 2022. Id. ¶ 64. DreamUp is a commercial product that relies on Stable Diffusion to produce images and is only available to customers who pay DeviantArt. Id. ¶¶ 35, 115. Plaintiffs allege that at least one LAION dataset that was incorporated into Stable Diffusion for training images (the “aesthetic dataset”) was procured by scraping primarily 100 websites, including DeviantArt's site. Id. ¶¶ 109, 110. As a result, plaintiffs allege that Stability copied thousands and possible millions of training images from DeviantArt created by artists and other DeviantArt subscribers without licensing their works of art. Id. ¶¶ 116-117. By incorporating DreamUp and therefore Stable Diffusion into its website, plaintiffs allege that DeviantArt is violating its own terms of service against using content for “commercial” purposes and without consent, as well as its privacy policy. This conduct, according to plaintiffs, represents “unfair competition against” DeviantArt's artist customers. Id. ¶¶ 123, 124.

Defendant Midjourney, based in San Francisco, created and distributes the “Midjourney Product.” Id. ¶¶ 34, 134. The Midjourney Product was launched in beta form in July 2022, and is alleged to be a commercial product that produces images in response to text prompts in the same manner as DreamStudio and DreamUp. Plaintiffs allege that the Midjourney product uses Stable Diffusion but also that it was “trained on a subset of the images used to train Stable Diffusion.” Id. ¶¶ 34, 62, 134, 135. The Midjourney Product is offered to online users of the internet-chat system Discord, as well as through an app, for a service fee. Id. ¶¶ 136, 137, 139. Midjourney's CEO has stated that Midjourney used large open data sets, thereby “implying” that Midjourney used the LAION datasets for training. Id. ¶¶ 148-149. In August 2022, Midjourney released a beta version using Stable Diffusion. Id. ¶ 149.

Plaintiff Anderson resides in Oregon and is a full-time cartoonist and illustrator. Plaintiffs allege that Anderson “has created and owns a copyright interest in over two hundred Works included in the Training Data,” and has registered or applied “for an owns copyright registrations for sixteen collections that include Works used as Training Images.” Id. ¶ 28. Plaintiff McKernan resides in Tennessee and is a full-time artist. McKernan is alleged to have “created and owns a copyright interest in over thirty Works used as Training Images.” Id. ¶ 20. Plaintiff Ortiz resides in California and is a full-time artist. Ortiz is alleged to have “created and owns a copyright interest in at least twelve Works that were used as Training Images.” Id. ¶ 30.

Plaintiffs assert the following claims against all three sets of defendants: (1) Direct Copyright Infringement, 17 U.S.C. § 106; (2) Vicarious Copyright Infringement, 17 U.S.C. § 106; (3) violation of the Digital Millennium Copyright Act, 17 U.S.C. §§ 1201-1205 (“DMCA”); (4) violation of the Right to Publicity, Cal. Civil Code § 3344; (5) violation of the Common Law Right of Publicity; (6) Unfair Competition, Cal. Bus. & Prof. Code § 17200; and (7) Declaratory Relief. Plaintiffs also assert a breach of contract claim against DeviantArt only.

Each defendant separately moves to dismiss, and DeviantArt also moves to strike under California's anti-SLAPP statute, California Code of Civil Procedure § 425.16 .

Plaintiffs have filed two notices of supplemental authority, Dkt. Nos. 112 & 115, and I have considered them both.

LEGAL STANDARD

Under FRCP 12(b)(6), a district court must dismiss a complaint if it fails to state a claim upon which relief can be granted. To survive a Rule 12(b)(6) motion to dismiss, the plaintiff must allege “enough facts to state a claim to relief that is plausible on its face.” Bell Atl. Corp. v. Twombly, 550 U.S. 544, 570 (2007). A claim is facially plausible when the plaintiff pleads facts that “allow the court to draw the reasonable inference that the defendant is liable for the misconduct alleged.” Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009) (citation omitted). There must be “more than a sheer possibility that a defendant has acted unlawfully.” Id. While courts do not require “heightened fact pleading of specifics,” a plaintiff must allege facts sufficient to “raise a right to relief above the speculative level.” Twombly, 550 U.S. at 555, 570.

In deciding whether the plaintiff has stated a claim upon which relief can be granted, the Court accepts the plaintiff's allegations as true and draws all reasonable inferences in favor of the plaintiff. See Usher v. City of Los Angeles, 828 F.2d 556, 561 (9th Cir. 1987). However, the court is not required to accept as true “allegations that are merely conclusory, unwarranted deductions of fact, or unreasonable inferences.” In re Gilead Scis. Sec. Litig., 536 F.3d 1049, 1055 (9th Cir. 2008). If the court dismisses the complaint, it “should grant leave to amend even if no request to amend the pleading was made, unless it determines that the pleading could not possibly be cured by the allegation of other facts.” Lopez v. Smith, 203 F.3d 1122, 1127 (9th Cir. 2000). In making this determination, the court should consider factors such as “the presence or absence of undue delay, bad faith, dilatory motive, repeated failure to cure deficiencies by previous amendments, undue prejudice to the opposing party and futility of the proposed amendment.” Moore v. Kayport Package Express, 885 F.2d 531, 538 (9th Cir. 1989).

DISCUSSION

I. MOTIONS TO DISMISS

Each of the defendants separately move to dismiss, but raise substantially similar arguments addressed collectively below.

In an overarching argument, DeviantArt complains that the Complaint contains undifferentiated allegations between the separate defendants and often lumps “defendants” together. DeviantArt Motion to Dismiss (Dkt. No. 49) at 7-8. It points out that group pleading makes it difficult for each defendant to figure out what exact conduct plaintiffs believe each defendant engaged in. Id. As discussed below, I am dismissing the majority of plaintiffs' claims but granting leave to amend. When plaintiffs amend, plaintiffs should not lump “defendants” together. Instead, they should identify each defendant by name with respect to conduct they allege each defendant engaged in.

A. Copyright - Counts I & II

1. Registration

Each defendant argues that McKernan and Ortiz's copyright claims must be dismissed because neither of them has registered their images with the Copyright Office. They also move to “limit” Anderson's copyright claim to infringement based only on the 16 collections of works that she has registered. See, e.g., Declaration of Paul M. Schoenhard (Dkt. No. 51-1), ¶¶ 5-6; see also Compl. ¶ 28 & Exs. 1-16.

Stability AI moves for judicial notice of the lack of copyright registrations. Plaintiffs do not oppose that judicial notice.

In opposition, plaintiffs do not address, much less contest, McKernan or Ortiz's asserted inability to pursue Copyright Act claims. At oral argument, plaintiffs' counsel clarified that they are not asserting copyright claims on behalf of these two plaintiffs. July 19, 2023 Transcript (Tr.), pg. 17:1-5. As such, McKernan and Ortiz's copyright act claims are DISMISSED WITH PREJUDICE.

Likewise, plaintiffs do not address or dispute that Anderson's copyright claims should be limited to the collections Anderson has registered. The scope of Anderson's Copyright Act claims are limited to the collections which she has registered.

2. Identifying Infringed Works

As to Anderson, defendants argue that she cannot proceed with her copyright infringement allegations unless she identifies with specificity each of her registered works that she believes were used as Training Images for Stable Diffusion. In the Complaint, Anderson alleges that she “has created and owns a copyright interest in over two hundred Works included in the Training Data” and that “[e]xamples of Ms. Andersen's Works included in the Training Data can be found here: https://haveibeentrained.com/?search text'sarah%20andersen.cites.” Compl., ¶ 28 & n.1.

Defendants contend that those allegations are insufficient and argue that Anderson should be required to identify which specific works from which of her registered collections she believes were copied into the LAION datasets and ended up as Training Images for Stable Diffusion. See, e.g., Stability Mot. at 4-5; DeviantArt Mot. to Dismiss (Dkt. No. 49) at 9; Midjourney Mot. (Dkt. No. 77) at 8-9.

Anderson does not identify which of her specific works covered by a registration were used as Training Images but relies on the output of a search of her name on the “ihavebeentrained.com” site to support the plausibility and reasonableness of her belief that her works were, in fact, used in the LAION datasets and training for Stable Diffusion. Compl., ¶ 28 & n.1. She attests that her review of the output pages from that search confirms that some of her registered works were used as Training Images. That is a sufficient basis to allow her copyright claims to proceed at this juncture, particularly in light of the nature of this case, i.e., that LAION scraped five billion images to create the Training Image datasets. At this juncture, the plausible inferences are that all of Anderson's works that were registered as collections and were online were scraped into the training datasets. Her assertions regarding the results of her search on the “haveibeentrained” site supports that inference and makes it reasonable for this case. While defendants complain that Anderson's reference to search results on the “haveibeentrained” website is insufficient, as the output pages show many hundreds of works that are not identified by specific artists, defendants may test Anderson's assertions in discovery.

The scope of the works scraped from the internet and allegedly used to create the Training Images distinguish this case from those defendants rely on. For example, Bespaq Corp. v. Haoshen Trading Co., No. C 04-3698 PJH, 2005 WL 14841, at *2 (N.D. Cal. Jan. 3, 2005), the copyright claims were dismissed with leave given the failure of the plaintiff to identify “which preexisting works in the registered catalog” containing pictures of miniature furniture “have been infringed by the defendants.” In Cutler v. Enzymes, Inc., No. C 08-04650 JF(RS), 2009 WL 482291, at *3 (N.D. Cal. Feb. 25, 2009), the case involved “only three copyrighted books” and “[a]side from claims of ownership, the complaint is devoid of any other specific facts related to the Published Work and alleged copyright infringement.”

See Declaration of Judd Lauter [Dkt. No. 52-1], Ex. A (screenshot of “haveibeentrained.com” website search referenced by Anderson containing hundreds of images).

3. Direct Infringement Allegations Against Stability

Plaintiffs' primary theory of direct copyright infringement is based on Stability's creation and use of “Training Images” scraped from the internet into the LAION datasets and then used to train Stable Diffusion. 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 (called “Training Images”) to train Stable Diffusion and caused those “images to be stored at and incorporated into Stable Diffusion as compressed copies.” Compl. ¶¶ 3-4, 25-26, 57. In its “Preliminary Statement” in support of its motion to dismiss, Stability opposes the truth of plaintiffs' assertions. See Stability Motion to Dismiss (Dkt. No. 58) at 1. However, 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. Id. Stability does not otherwise oppose the sufficiency of the allegations supporting Anderson's direct copyright infringement claims with respect to the Training Images.

Stability's motion to dismiss Count I for direct copyright infringement is DENIED.

4. Direct Infringement Allegations Against DeviantArt

Plaintiffs fail to allege specific plausible facts that DeviantArt played any affirmative role in the scraping and using of Anderson's and other's registered works to create the Training Images. The Complaint, instead, admits that the scraping and creation of Training Images was done by LAION at the direction of Stability and that Stability used the Training Images to train Stable Diffusion. Compl. ¶¶ 2-4, 104-107. What DeviantArt is specifically alleged to have done is be a primary “source” for the “LAION-Aesthetic dataset” created to train Stable Diffusion. Id. ¶¶ 109-110, 116. That, however, does not support a claim of direct copyright infringement by DeviantArt itself.

In opposition, plaintiffs offer three theories of DeviantArt's direct infringement:

(1) direct infringement by distributing Stable Diffusion, which contains compressed copies of the training images, as part of DeviantArt's DreamUp AI imaging product; (2) direct infringement by creating and distributing their DreamUp, which is itself an infringing derivative work; and (3) generating and distributing output images which are infringing derivative works.
DDkt. No. 65 at 6. In support, plaintiffs point to their allegations that: “Stable Diffusion has been used as a Software Library within” DreamUp, Id. ¶ 23; “DreamUp is a commercial product that relies on Stable Diffusion to produce images,” Id. ¶ 35; “DreamUp is a web-based app that generates images in response to Text Prompts. Like DreamStudio, DreamUp relies on Stability's Stable Diffusion software as its underlying software engine,” Id. ¶ 64; that DeviantArt embraced “Stable Diffusion by incorporating it into their website via the DreamUp app,” Id. ¶ 123; and DeviantArt decided to use “‘Stable Diffusion because it's the only option for us to take an open source [software engine] and modify it.'” Id. ¶ 129 (quoting DeviantArt CEO).

DeviantArt vigorously disputes the assertions - made throughout the Complaint - that “embedded and stored compressed copies of the Training Images” are contained within Stable Diffusion. See Compl. ¶¶ 3, 19, 58; see also ¶ 18 (Output Images are derivative works as they contain “contain compressed copies of the copyrighted works they were trained on.”). DeviantArt (and Stability and Midjourney) argue that those assertions are implausible given plaintiffs' allegation that the training dataset was comprised of five billion images; five billion images could not possibly be compressed into an active program. Defendants also claim that the “compressed copies” allegations are contradicted by plaintiffs' descriptions of the diffusion process in the Complaint. Those descriptions admit that the diffusion process involves not copying of images, but instead the application of mathematical equations and algorithms to capture concepts from the Training Images. Compl. ¶¶ 70, 75, 79. Finally, defendants rely heavily on plaintiffs' admission that, “[i]n general, none of the Stable Diffusion output images provided in response to a particular Text Prompt is likely to be a close match for any specific image in the training data.” Id. ¶ 93. In light of that, defendants argue that plaintiffs cannot plausibly plead copying in violation of the Copyright Act based on Output Images.

Turning to the first theory of direct copyright infringement and the plausibility of plaintiffs' assertion that Stable Diffusion contains “compressed copies” of the Training Images and DeviantArt's DreamUp product utilizes those compress copies, DeviantArt is correct that the Complaint is unclear. As noted above, the Complaint repeatedly alleges that Stable Diffusion contains compressed copies of registered works. But the Complaint also describes the diffusion practice as follows:

Because a trained diffusion model can produce a copy of any of its Training Images-which could number in the billions-the diffusion model can be considered an alternative way of storing a copy of those images. In essence, it's similar to having a directory on your computer of billions of JPEG image files. But the diffusion model uses statistical and mathematical methods to store these images in an even more efficient and compressed manner.
Compl. ¶ 75(c).

Plaintiffs will be required to amend to clarify their theory with respect to compressed copies of Training Images and to state facts in support of how Stable Diffusion - a program that is open source, at least in part - operates with respect to the Training Images. If plaintiffs contend Stable Diffusion contains “compressed copies” of the Training Images, they need to define “compressed copies” and explain plausible facts in support. And if plaintiffs' compressed copies theory is based on a contention that Stable Diffusion contains mathematical or statistical methods that can be carried out through algorithms or instructions in order to reconstruct the Training Images in whole or in part to create the new Output Images, they need to clarify that and provide plausible facts in support.

At the hearing, the parties disputed how much information plaintiffs could discover from reviewing the open source and or publicly available code for Stable Diffusion. Plaintiffs asserted that in order to really figure out how Training Images are present in Stable Diffusion, they would need access to information that has not been publicly released by Stability. Stability responded that the only information that may not be publicly available are “weight files” that allow users like DeviantArt and Midjourney to set different weights and parameters for their output, but are irrelevant to the “compressed copies” issue. Compare Tr. at 6:13-20 with Tr. at 28:1-11. On amendment, plaintiffs might consider pleading facts regarding what they can or cannot determine from review of the Stable Diffusion code.

Depending on the facts alleged on amendment, DeviantArt (and Midjourney) may make a more targeted attack on the direct infringement contentions. It is unclear, for example, if Stable Diffusion contains only algorithms and instructions that can be applied to the creation of images that include only a few elements of a copyrighted Training Image, whether DeviantArt or Midjourney can be liable for direct infringement by offering their clients use of the Stable Diffusion “library” through their own apps and websites. But if plaintiffs can plausibly plead that defendants' AI products allow users to create new works by expressly referencing Anderson's works by name, the inferences about how and how much of Anderson's protected content remains in Stable Diffusion or is used by the AI end-products might be stronger.

Plaintiffs' second theory of direct infringement - that Stable Diffusion is a “derivative work” because it contains compressed copies of billions of copyrighted images and by incorporating Stable Diffusion into DreamUp, DeviantArt is liable for producing works that have been “transformed” based on plaintiffs' works, Dkt. No. 65 at 13-14 - fails for the same reasons.

In addition to providing clarity regarding their definition of and theory with respect to the inclusion of compressed copies of Training Images in Stable Diffusion, plaintiffs shall also provide more facts that plausibly show how DeviantArt is liable for direct copyright infringement when, according to plaintiffs' current allegations, DeviantArt simply provides its customers access to Stable Diffusion as a library. Plaintiffs do cite testimony from DeviantArt's CEO that DeviantArt uses Stable Diffusion because Stability allowed DeviantArt to “modify” Stable Diffusion. Compl. ¶ 129. The problem is that there are no allegations what those modifications might be or why, given the structure of Stable Diffusion, any compressed copies of copyrighted works that may be present in Stable Diffusion would be copied within the meaning of the Copyright Act by DeviantArt or its users when they use DreamUp. Nor do plaintiffs provide plausible facts regarding DeviantArt “distributing” Stable Diffusion to its users when users access DreamUp through the app or through DeviantArt's website.

That leaves plaintiffs' third theory of direct infringement; that DreamUp produces “Output Images” that are all infringing derivative works. DeviantArt argues that to adequately plead this claim, plaintiffs must allege the Output Images are substantially similar to the protected works but they cannot do so given plaintiffs' repeated admission that “none of the Stable Diffusion output images provided in response to a particular Text Prompt is likely to be a close match for any specific image in the training data.” Compl. ¶ 93; see Skidmore as Tr. for Randy Craig Wolfe Tr. v. Led Zeppelin, 952 F.3d 1051, 1064 (9th Cir. 2020 (“the hallmark of ‘unlawful appropriation' is that the works share substantial similarities.”). Plaintiffs argue that they do not need to plead or address substantial similarity under Range Rd. Music, Inc. v. E. Coast Foods, Inc., 668 F.3d 1148 (9th Cir. 2012). There, addressing allegations that copyrighted music was played from a compact disc and performed live, the court held:

See also 17 U.S.C. § 101: “A ‘derivative work' is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted. A work consisting of editorial revisions, annotations, elaborations, or other modifications which, as a whole, represent an original work of authorship, is a ‘derivative work'.”

See also id. (“In our circuit, we use a two-part test to determine whether the defendant's work is substantially similar to the plaintiff's copyrighted work. [] The first part, the extrinsic test, compares the objective similarities of specific expressive elements in the two works ....The second part, the intrinsic test, “test[s] for similarity of expression from the standpoint of the ordinary reasonable observer, with no expert assistance.” (quoting, Jada Toys, Inc. v. Mattel, Inc., 518 F.3d 628, 637 (9th Cir. 2008)) other internal citations omitted).

A showing of “substantial similarity” is irrelevant in a case like this one, in which the Music Companies produced evidence that the public performances entailed direct copying of copyrighted works. []. (noting that a demonstration of substantial similarity is only necessary to prove infringement “[a]bsent evidence of direct copying”); see also Narell v. Freeman, 872 F.2d 907, 910 (9th Cir.1989) (noting that “[a] finding that a defendant copied a plaintiff's work, without application of a substantial similarity analysis” will be made “when the defendant has engaged in virtual duplication of a plaintiff's entire work”); 2 Howard B. Abrams, The Law of Copyright § 14:10 (2011) (“Direct proof [of copying] can consist of ... testimony of direct observation of the infringing act....”).
Id. at 1154. Plaintiffs rely on that line of cases and point to their allegation that all elements of plaintiff Anderson's copyrighted works (and the copyrighted works of all others in the purported class) were copied wholesale as Training Images and therefore the Output Images are necessarily derivative. See Compl. ¶ 95 (“Every output image from the system is derived exclusively from the latent images, which are copies of copyrighted images. For these reasons, every hybrid image is necessarily a derivative work.”).

A problem for plaintiffs is that unlike in Range Road - observed wholesale copying and performing - the theory regarding compressed copies and DeviantArt's copying need to be clarified and adequately supported by plausible facts. See supra. The other problem for plaintiffs is that it is simply not plausible that every Training Image used to train Stable Diffusion was copyrighted (as opposed to copyrightable), or that all DeviantArt users' Output Images rely upon (theoretically) copyrighted Training Images, and therefore all Output images are derivative images.

Even if that clarity is provided and even if plaintiffs narrow their allegations to limit them to Output Images that draw upon Training Images based upon copyrighted images, I am not convinced that copyright claims based a derivative theory can survive absent “substantial similarity” type allegations. The cases plaintiffs rely on appear to recognize that the alleged infringer's derivative work must still bear some similarity to the original work or contain the protected elements of the original work. See, e.g., Jarvis v. K2 Inc., 486 F.3d 526, 532 (9th Cir. 2007) (finding works were derivative where plaintiff “delivered the images to K2 in one form, and they were subsequently used in the collage ads in a quite different (though still recognizable) form. The ads did not simply compile or collect Jarvis' images but rather altered them in various ways and fused them with other images and artistic elements into new works that were based on- i.e., derivative of-Jarvis' original images.”) (emphasis added); ITC Textile Ltd. v. Wal-Mart Stores Inc., No. CV122650JFWAJWX, 2015 WL 12712311, at *5 (C.D. Cal. Dec. 16, 2015) (“Accordingly, even if Defendants did modify them slightly, such modifications are not sufficient to avoid infringement in a direct copying case.... Thus, the law is clear that in cases of direct copying, the fact that the final result of defendant's work differs from plaintiff's work is not exonerating.”) (emphasis added); see also Litchfield v. Spielberg, 736 F.2d 1352, 1357 (9th Cir. 1984) (“a work is not derivative unless it has been substantially copied from the prior work”); Authors Guild v. Google, Inc., 804 F.3d 202, 225 (2d Cir. 2015) (“derivative works over which the author of the original enjoys exclusive rights ordinarily are those that re-present the protected aspects of the original work, i.e., its expressive content”).

Defendants make a strong case that I should dismiss the derivative work theory without leave to amend because plaintiffs cannot plausibly allege the Output Images are substantially similar or re-present protected aspects of copyrighted Training Images, especially in light of plaintiffs' admission that Output Images are unlikely to look like the Training Images. Compl. ¶ 93. But other parts of plaintiffs' Complaint allege that Output Images can be so similar to plaintiff's styles or artistic identities to be misconstrued as “fakes.” Compl. ¶¶ 44ii, 171-172. Once plaintiffs amend, hopefully providing clarified theories and plausible facts, this argument may be re-raised on a subsequent motion to dismiss.

DeviantArt's motion to dismiss Claim I is GRANTED with leave to amend.

5. Direct Infringement Contentions Against Midjourney

Plaintiffs allege the following regarding the Midjourney product: “the Midjourney Product is a commercial product that produces images in response to text prompts. On information and belief, Stable Diffusion was used in iterations of the Midjourney Product. On information and belief, the version of the Midjourney Product currently available was trained on a subset of the images used to train Stable Diffusion,” Compl. ¶ 34; “Midjourney's main product is an online AI-based image generator offered under the name “Midjourney.” Like DreamUp and DreamStudio, the Midjourney image generator uses Text Prompts as input and produces digital images as output. Just like DreamUp and DreamStudio, Midjourney relies on Stable Diffusion as its underlying software engine for generating images,” Id. ¶ 134; “Midjourney subscribers also receive access to the Midjourney web app, similar to DreamStudio or DreamUp, which lets users access the Midjourney service through a web interface,” Id. ¶ 139; “Midjourney relies on appropriating millions of copyrighted images created by artists and using these images as Training Images,” Id. ¶ 144; “Midjourney is a collage tool, only capable of producing images that are remixed and reassembled from the copyrighted work of others,” Id. ¶ 145; Midjourney's CEO, when asked how datasets Midjourney uses replied, “[i]t's just a big scrape of the internet. We use the open data sets that are published and train across those,” Id. ¶ 148.

In opposition to Midjourney's motion to dismiss, plaintiffs identify theories of direct infringement against Midjourney that differ slightly from those offered against DeviantArt:

(1) direct infringement by reproducing protected works by scraping (i.e., copying) and reproducing images used as training images for the Midjourney Product; (2) direct infringement by distributing Stable Diffusion, which contains compressed copies of the training images, as part of the Midjourney Product; (3) direct infringement by creating and distributing their Midjourney Product, which is itself an infringing derivative work; and (4) generating and distributing output images which are infringing derivative works.
Dkt. No. 67 at 5. On the first theory, unlike the detailed allegations regarding Stability's role with LAION and Stability's connection to the scraping and then use of the Training Images for Stable Diffusion, there are no facts regarding what training, if any, Midjourney conducted for its Midjourney product. The only reference is to the comments of Midjourney's CEO that Midjourney uses the open datasets (that plaintiffs presume are from LAION, Compl. ¶ 149) and they and everyone else “train” across them. Id. ¶ 148. However, plaintiffs also allege that Midjourney uses Stable Diffusion. Id. ¶ 23 (“Stable Diffusion has been used as a Software Library within multiple programs, including DreamStudio, DreamUp, and, on information and belief, the Midjourney Product.”); see also ¶ 17 (“‘Work' or ‘Works' refers to any image that was used to train any version of Stable Diffusion that was offered directly and/or incorporated into another product by one or more Defendants during the Class Period.”). 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?

With respect to the other theories, each of them fail for the reasons identified above with respect to DeviantArt. Midjourney's motion to dismiss Claim 1 is GRANTED with leave to amend.

In its motion and reasserted during the hearing on these motions, Midjourney argued leave to amend should not be granted for plaintiffs' claim of infringement by performance in violation of 17 U.S.C. § 106(4), because there is no performance right in pictorial works and plaintiffs did not address or oppose dismissal in their opposition. Compl. ¶ 160(d); Tr. at 22-23. At the hearing, plaintiffs requested leave to allege their theory of performance in violation of section 106(4). Tr. at 32:16-19. Leave is granted to allege all theories of copyright infringement. However, I will not be as generous with leave to amend on the next, expected rounds of motions to dismiss and I will expect a greater level of specificity as to each claim alleged and the conduct of each defendant in support of each claim.

6. Vicarious Infringement

Because plaintiffs have not alleged claims of direct infringement against DeviantArt or Midjourney, the vicarious infringement claims cannot be sustained against them. See Perfect 10, Inc. v. Yandex N.V., 962 F.Supp.2d 1146, 1158 (N.D. Cal. 2013), as amended (Sept. 6, 2013) (“vicarious liability requires an underlying act of direct infringement”). Claim 2 is DISMISSED against DeviantArt and Midjourney with leave to amend.

Midjourney also argues that the vicarious liability claim should be dismissed because “a defendant cannot be secondarily liable for their own direct infringement.” Sound & Color, LLC v. Smith, No. 222CV01508ABASX, 2023 WL 2821881, at *16 (C.D. Cal. Feb. 28, 2023). However, only after the court determines what, if any claims, for direct infringement are adequately alleged against each defendant, can the court determine whether plaintiffs are attempting to hold a defendant secondarily liable for their own direct infringement or instead hold them liable for others' infringement when using defendants' generative AI products.

Plaintiffs have adequately alleged direct infringement against Stability AI given their allegations regarding Stability's involvement in the scraping, copying, and use of Training Images to train Stable Diffusion. However, to be liable for vicarious copyright infringement, “a plaintiff must allege that the defendant has (1) the right and ability to supervise the infringing conduct and (2) a direct financial interest in the infringing activity.” Perfect 10, Inc. v. Visa Int'l Serv. Ass'n, 494 F.3d 788, 802 (9th Cir. 2007). Plaintiffs have been given leave to amend to clarify their theory and add plausible facts regarding “compressed copies” in Stable Diffusion and how those copies are present (in a manner that violates the rights protected by the Copyright Act) in or invoked by the DreamStudio, DreamUp, and Midjourney products offered to third parties. That same clarity and plausible allegations must be offered to potentially hold Stability vicariously liable for the use of its product, DreamStudio, by third parties. Compl. ¶ 44ii. (defining common question as “[w]hether Defendants vicariously violated the copyrights of Plaintiffs and the Class when third parties used Defendants' products to create Fakes”); Id. ¶¶ 169-177 (vicarious liability based on third party “imposters” creating “fakes” using defendants' products).

I recognize that if plaintiffs are able to clarify their theory and add plausible facts that copyrighted compressed copies used or invoked by defendants' products, there are other issues with plaintiffs' vicarious infringement theory based on “imposters” using defendants' tools to produce “fakes.” Plaintiffs' Complaint is devoid of any allegation that any of Anderson's works (or any other class member) were used to create “fakes” of their works. Instead, the Complaint alleges that based on how the diffusion process works, none of the Output Images are likely “to be a close match for any specific image in the training data.” Compl. ¶ 93. In opposition, plaintiffs appear to back away from reliance on the imposters and fakes assertions, and instead rely on their theory that all Output Images are derivative infringing works. See Dkt. No. 65 at 11; Dkt. No. 66 at 11; Dkt. No. 67 at 10. The deficiencies with that theory have been identified above.

Count II is DISMISSED as to each defendant with leave to amend.

B. DMCA

Section 1202(b) of the Digital Millennium Copyright Act (“DMCA”) provides that:

No person shall, without the authority of the copyright owner or the law-
1. intentionally remove or alter any copyright management information;
2. distribute or import for distribution any copyright management information knowing that the copyright management information has been removed or altered without authority of the copyright owner . . . or
3. distribute, import for distribution . . . works [or] copies of works . . . knowing that copyright management information has been removed or altered without authority of the copyright owner . . .
knowing or having reasonable grounds to know that it will induce, enable, facilitate or conceal an infringement of any right under this title.
17 U.S.C. § 1202(b).

Section 1202(c) defines copyright management information (CMI) to include the following: “[the] title and other information identifying the work, including the information set forth on a notice of copyright;” “[the] name of, and other identifying information about, the author of a work;” and “[the] name of, and other identifying information about, the copyright owner of the work, including the information set forth in a notice of copyright.” 17 U.S.C. § 1202(c).

In the Ninth Circuit, to establish knowing or having reasonable grounds to know that conduct will “induce, enable, facilitate or conceal,” a plaintiff “must make an affirmative showing, such as by demonstrating a past ‘pattern of conduct' or ‘modus operandi,' that the defendant was aware or had reasonable grounds to be aware of the probable future impact of its actions.” Stevens v. Corelogic, Inc., 899 F.3d 666, 674 (9th Cir. 2018). At the summary judgment stage, this claim requires providing evidence that the alleged infringer knew that the removal of the CMI would “induce, enable, facilitate, or conceal” copyright infringement. Id. at 673. At the pleading stage, the claimant must plead facts plausibly showing that the alleged infringer had this required mental state. Philpot v. Alternet Media, Inc., No. 18-cv-04479-TSH, 2018 WL 6267876, at *5 (N.D. Cal. Nov. 30, 2018). While “[a]t the pleading stage, mental conditions generally need not be alleged with specificity,” a plaintiff must still “allege sufficient facts to support the reasonable inference that the defendant ‘knew or had a reasonable basis to know that the removal or alteration of CMI ... w[ould] aid infringement.'” Doe 1 v. GitHub, Inc., No. 22-CV-06823-JST, 2023 WL 3449131, at *12 (N.D. Cal. May 11, 2023) (quoting Harrison v. Pinterest, Inc., No. 20-cv-05290-EJD, 2022 WL 4348460, at * 5 (N.D. Cal. Sept. 19, 2022)).

Stability moves to dismiss the DMCA claim because plaintiffs fail to allege that Stability removed any “copyright management information” (“CMI”) from any particular work of the plaintiffs. It contends that there are no allegations in the Complaint that any of the plaintiffs included identified CMI in particular works that were available online, or facts plausibly showing that when the images were scraped and included in training datasets plaintiffs' CMI was removed. Finally, it claims that there are no facts alleged that could plausibly show it - as opposed to LAION or others- had the requisite “double-scienter”: in other words, facts plausibly supporting that each defendant knew CMI was being scraped from plaintiffs' works and knowing that conduct would “induce, enable, facilitate, or conceal an infringement.” 17 U.S.C. § 1202(b); Stevens, 899 F.3d at 674-75 (“the ‘induce, enable, facilitate or conceal' requirement is intended to limit liability in some fashion - specifically, to instances in which the defendant knows or has a reasonable basis to know that the removal or alteration of CMI or the distribution of works with CMI removed will aid infringement.”). DeviantArt and Midjourney make similar arguments, albeit from a stronger position because there are no allegations that either of them was involved with LAION or directly with the training of Stable Diffusion where, presumably, the removal or alteration of the CMI occurred.

Defendants also allege that plaintiffs fail to allege facts plausibly supporting that Stability or some other entity (e.g., LAION) scraped any of their works into their training dataset. Similar to the discussions above, in connection with Anderson's copyright claim, given the particular facts of this case and the allegation that five billion works were scraped into the LAION datasets, plaintiffs' reliance on the “haveibeentrained” site and their assertions that their works show up on searches on that site are sufficient to plausibly allege their works were scraped into the LAION datasets. Compl. ¶¶ 28-30 & fns. 1-3.

In response, plaintiffs point to paragraphs180 and 191 of their Complaint, where they allege generally that plaintiffs and “others” in the putative class included various categories of CMI in their works and the “removal or alteration” of that CMI by defendants, including “the creator's name” and “the form of artist's signatures.” These allegations are wholly conclusory. In order to state this claim, each plaintiff must identify the exact type of CMI included in their online works that were online and that they have a good faith belief were scraped into the LAION datasets or other datasets used to train Stable Diffusion. At the hearing, plaintiffs argued that it is key for the development of generative AI models to capture not only images but any accompanying text because that accompanying text is necessary to the models' ability to “train” on key words associated with those images. Tr. at 9:13-24. But there is nothing in the Complaint about text CMI present in the images the named plaintiffs included with their online images that they contend was stripped or altered in violation of the DMCA during the training of Stable Diffusion or the use of the end-products. Plaintiffs must, on amendment, identify the particular types of their CMI from their works that they believe were removed or altered.

In addition, plaintiffs must clarify and then allege plausible facts regarding which defendants they contend did the stripping or altering in violation of the DMCA and when that occurred. The Complaint pleads facts which put the responsibility for the initial scrapping of images on LAION and/or Stability. Compl. ¶¶ 102-110. While Complaint attempts to place responsibility for stripping or altering of CMI at the time when the “Defendants” “trained Stable Diffusion,” Id. ¶ 183, there are no facts at all regarding DeviantArt's or Midjourney's training of Stable Diffusion.

The DMCA claim is DISMISSED as to each defendant with leave to amend. Plaintiffs shall identify the specific CMI each named plaintiff included in the images that each plaintiff contends was used to training Stable Diffusion. Plaintiffs shall not allege violations of the DMCA by “defendants,” but shall instead identify with specificity the theory of DMCA liability for each defendant and plausible facts in support with respect to each defendant.

C. Right of Publicity Claims

1. Failure to State the Claims

Plaintiffs allege two species of right of publicity claims. In support of their statutory claim (under Cal. Civ. Code § 3344), they assert that defendants “knowingly” used plaintiffs' names in their products - by allowing users to request art in the style of their names - and that their names are uniquely associated with their art and distinctive artistic styles. Compl. ¶ 204. In support of their common law claim, they allege that defendants are violating their rights in their “artistic identities” - not just their rights in their works - because the products allow users to request Output Images “in the style” of their artistic identities. Id. ¶¶ 214-222.

See Downing v. Abercrombie & Fitch, 265 F.3d 994, 1001 (9th Cir. 2001) (“California has long recognized a common law right of privacy for protection of a person's name and likeness against appropriation by others for their advantage. [] To sustain a common law cause of action for commercial misappropriation, a plaintiff must prove: ‘(1) the defendant's use of the plaintiff's identity; (2) the appropriation of plaintiff's name or likeness to defendant's advantage, commercially or otherwise; (3) lack of consent; and (4) resulting injury.' ....In addition to the common law cause of action, California has provided a statutory remedy for commercial misappropriation under California Civil Code § 3344....Under section 3344, a plaintiff must prove all the elements of the common law cause of action. In addition, the plaintiff must allege a knowing use by the defendant as well as a direct connection between the alleged use and the commercial purpose.”) (quoting Eastwood v. Superior Court, 149 Cal.App.3d 409, 417 (1983).

In their opposition briefs, however, plaintiffs retreat from reliance on their “artistic identities” or distinctive styles allegations and recast both types of right of publicity claims as based on each defendant's “misuse of their names” by associating their names with AI imaging output for defendants' “commercial purposes.” See, e.g., Dkt. No. 65 at 18-21; Dkt. No. 66 at 1923; Dkt. No. 67 at 18-20. At the hearing, plaintiffs provided further clarification; that both claims are based on defendants' use of plaintiffs' names to advertise and promote their DreamStudio, DreamUp, and Midjourney products. Tr. at 10: 15-20; see also Compl. ¶ 203 (“Defendants appropriated Plaintiffs' names to Defendants' advantage, including for the purposes of advertising, selling, and soliciting purchases through Defendants' AI Image Products.”); Id. ¶ 205 (“There is a direct connection between Defendants' misappropriation of Plaintiffs' names and Defendants' commercial purposes, because Defendants used Plaintiffs' names to advertise art “in the style” of Plaintiffs' work. Defendants used Plaintiffs' names and advertised their AI's ability to copy or generate work in the artistic style that Plaintiffs popularized in order to sell Defendants' products and services. Defendants' ability to market art similar to and associated with Plaintiffs' names also enabled Defendants to establish an advantage over actual and prospective competitors.”).

The problem for plaintiffs is that nowhere in the Complaint have they provided any facts specific to the three named plaintiffs to plausibly allege that any defendant has used a named plaintiff's name to advertise, sell, or solicit purchase of DreamStudio, DreamUp or the Midjourney product. Nor are there any allegations regarding how use of these plaintiffs' names in the products' text prompts would produce an “AI-generated image similar enough that people familiar with Plaintiffs' artistic style could believe that Plaintiffs created the image,” and result in plausible harm to their goodwill associated with their names, in light of the arguably contradictory allegation that none of the Output Images are likely to be a “close match” for any of the Training Images. Compare Dkt. No. 66 at 21 with Compl. ¶ 93. Plaintiffs need to clarify their right of publicity theories as well as allege plausible facts in support regarding each defendants' use of each plaintiffs' name in connection with advertising specifically and any other commercial interests of defendants.

Plaintiffs disclaim any right to publicity claim based on users of DreamUp and the Midjourney products inputting their names into the text prompts. Dkt. No. 65 at 19 (“Plaintiffs do not challenge the use of names in prompts after the AI Image Product's release, but DeviantArt's intentional decision to reference Plaintiffs' identities by their name in DreamUp's prompts and use Plaintiffs' names to advertise and gain a competitive advantage for DreamUp”); Dkt. No. 67 (“Plaintiffs do not claim that the use of Plaintiffs' names by Midjourney users in text prompts is the basis for its right of publicity claims. Rather, Plaintiffs allege that Midjourney designed its AI Image Products to respond to the names of any artist included in its training data.”).

Plaintiffs' right of publicity claims are DISMISSED with leave to amend.

The parties spent much time in their briefs arguing whether the right of publicity claims are preempted because they seek to protect the same rights as the Copyright Act. Defendants made a strong showing of preemption, to the extent the Complaint based the publicity claims on the artistic “styles” of plaintiffs. However, once plaintiffs amend to clarify and presumably limit the right of publicity claims to the use of their names, with plausible facts in support, defendants will be able to make a more targeted preemption argument and the court may more readily determine whether plaintiffs' alleged use of their “names” is simply a preempted attempt to protect plaintiffs' copyrightable pictorial works of art or a not-preempted attempt to protect their names and goodwill.

2. First Amendment Defense

DeviantArt separately moves to dismiss (and strike) plaintiffs' right of publicity claim, contending that because plaintiffs are challenging DeviantArt's expressive conduct - the creation of new artistic works - the First Amendment demands the balancing of plaintiffs' publicity rights against the right of free expression by considering the “transformative use” of DreamUp's output.

Application of that test precludes this claim, according to DeviantArt, because plaintiffs admit that DreamUp relies on insights and interpolations from billions of images and directions from the user to produce a new work with a different purpose and different character.

Well-established law acknowledges “transformative use” as a defense to a right of publicity claim.

The California Supreme Court formulated the transformative use defense in Comedy III Productions, Inc. v. Gary Saderup, Inc., 25 Cal.4th 387 [] (2001). The defense is “a balancing test between the First Amendment and the right of publicity based on whether the work in question adds significant creative elements so as to be transformed into something more than a mere celebrity likeness or imitation.” Id. []. The California Supreme Court explained that “when a work contains significant transformative elements, it is not only especially worthy of First Amendment protection, but it is also less likely to interfere with the economic interest protected by the right of publicity.” Id. []. The court rejected the wholesale importation of the copyright “fair use” defense into right-of-publicity claims, but recognized that some aspects of that defense are “particularly pertinent.” Id.; see 17 U.S.C. § 107; see also SOFA Entm't, Inc. v. Dodger Prods., Inc., 709 F.3d 1273, 1277-78 (9th Cir.2013) (discussing the “fair use” defense codified in 17 U.S.C. § 107).
In re NCAA Student-Athlete Name & Likeness Licensing Litig., 724 F.3d 1268, 1273-74 (9th Cir. 2013).

Courts consider “at least” five factors to consider when determining if a work is “substantially transformative.” In re NCAA Student-Athlete Name & Likeness Licensing Litig., 724 F.3d 1268, 1274 (9th Cir. 2013). “First, if ‘the celebrity likeness is one of the ‘raw materials' from which an original work is synthesized,' it is more likely to be transformative than if ‘the depiction or imitation of the celebrity is the very sum and substance of the work in question.'” (quoting Comedy III Productions, Inc. v. Gary Saderup, Inc., 25 Cal.4th 387, 406 (2001)). Second, “the work is protected if it is ‘primarily the defendant's own expression'-as long as that expression is ‘something other than the likeness of the celebrity.'” Id. “Third, to avoid making judgments concerning ‘the quality of the artistic contribution,' a court should conduct an inquiry ‘more quantitative than qualitative' and ask ‘whether the literal and imitative or the creative elements predominate in the work.” Comedy III, 106 Cal.Rptr.2d 126, 21 P.3d at 809. Fourth, the California Supreme Court recognized that “a subsidiary inquiry” would be useful in close cases: whether “the marketability and economic value of the challenged work derive primarily from the fame of the celebrity depicted.” Id.,106 Cal.Rptr.2d 126. Finally, the court explained that “when an artist's skill and talent is manifestly subordinated to the overall goal of creating a conventional portrait of a celebrity so as to commercially exploit his or her fame,” the work is not transformative. Id.

Plaintiffs argue that it is inappropriate to consider the transformative use defense at the motion to dismiss stage. See id at 1274; see also Hilton v. Hallmark Cards, 599 F.3d 894, 910 (9th Cir. 2010) (“The applicability of the defense, however, does not preclude Hilton from showing the ‘minimal merit' needed to defeat Hallmark's motion to strike. Only if Hallmark is entitled to the defense as a matter of law can it prevail on its motion to strike. In this context, we note [] the application of the defense as a question of fact.”). They point out that how much of the style of an artist invoked by name is used in DreamUp's Output Images is a factual dispute. There are also disputes over whether users of DreamUp are motivated to obtain a reproduction of an artists' work or the DreamUp's transformative work. Finally, they assert that their claim is based (to be further clarified on amendment) on defendants' misappropriation of their names and use of their names to market their product and make sales.

I agree that the applicability of transformative use defense is better determined after plaintiffs clarify and otherwise amend their right of publicity claims and at a subsequent juncture on an evidentiary basis. While it is true that plaintiffs' current Complaint appears to admit that the DeviantArt's Output Images are not likely to be substantially similar to plaintiffs' works captured as Training Images, and therefore may be the result of substantial transformation, how the transformative use defense applies to works based on prompts of specific artists' names remains to be seen and should be tested on an evidentiary record. DeviantArt may raise this defense again once plaintiffs have amended their complaint and clarified their theories of liability for the right to publicity claims.

D. UCL

Defendants move to dismiss plaintiffs' unfair competition claim (“UCL”). That claim is asserted under the Lanham Act, under the common law, and under California Business & Professions Code section 17200, including unlawful prong claims based on copyright infringement and violation of the DMCA. Compl. ¶¶ 223-226.

As an initial matter, plaintiffs cannot tie their unlawful prong UCL claim to purported copyright violations. Those claims are preempted by the Copyright Act. “To the extent the improper business act complained of is based on copyright infringement, the claim was properly dismissed because it is preempted.” Sybersound Recs., Inc. v. UAV Corp., 517 F.3d 1137, 1152 (9th Cir. 2008) (dismissing UCL claim based on copyright infringement).

Plaintiffs' unlawful prong claim based on violations of the DMCA is dismissed with leave, as the underlying DMCA claim is dismissed with leave.

Defendants also argue the UCL claim must be dismissed to the extent it is based on the Lanham Act because plaintiffs' allegations admit they cannot allege a likelihood of confusion with respect to the Output Images. However, plaintiffs clarify in their opposition briefs that their Lanham Act/UCL claim is based on deception as to the “origin, sponsorship, or approval” of the works by the plaintiffs. See, e.g., Dkt. No. 65 at 23-24. That theory appears nowhere in the Complaint nor are there plausible facts alleged in support of how a user could be deceived that one of the named plaintiffs was the origin of an Output Image, sponsored the Output Image, or approved of the Output Images such that their goodwill was injured or they suffered other specific injury. Plaintiffs are given leave to amend to reallege the Lanham Act/UCL claim, pleading plausible facts that defendants' products create deception as to the origin or sponsorship of an output work that harms the named plaintiffs.

Plaintiffs contend that they have stated unfair and fraudulent prong claims under the UCL based on defendants' “unfair” and fraudulent misappropriation and copying of their art for commercial gain without permission or attribution in a manner likely to deceive the public. They contend that this “use” claim is different from the illegal UCL prong claim and not preempted by the Copyright Act. However, the unfair and fraudulent prong UCL claims must be dismissed. That dismissal is with leave to amend so that plaintiffs can allege facts regarding how each defendants' AI product uses plaintiff's names or associates works with plaintiffs and, for purposes of any claim under the fraudulent prong, facts that meet the heightened Rule 9(b) pleading standard.

Finally, plaintiffs assert they have pleaded a “common law” UCL claim, based on use of their names and use of their art as property. “The common law tort of unfair competition is generally thought to be synonymous with the act of ‘passing off' one's goods as those of another.” Bank of the W. v. Superior Court, 2 Cal.4th 1254, 1263 (1992) (explaining that the tort provided “an equitable remedy against the wrongful exploitation of trade names and common law trademarks that were not otherwise entitled to legal protection”); see also Sybersound Recs., Inc. v. UAV Corp., 517 F.3d 1137, 1153 (9th Cir. 2008). Given the shifting nature of plaintiffs' claims -from misappropriation of their styles or artistic identities to misappropriation of their names and associating their names with works not their own - the common law UCL claim is likewise dismissed, but with leave to amend to allow plaintiffs to clarify their theory and plead facts that could plausibly support a passing off claim.

The tort of unfair competition under California law requires plaintiffs to prove: (1) that plaintiffs have invested substantial time, skill or money in developing their property; (2) that defendants appropriated and used that property at little or no cost; (3) that defendants' appropriation and use of plaintiffs' property was without the authorization or consent; and (4) that plaintiffs were injured. See City Sols., Inc. v. Clear Channel Commc'ns, 365 F.3d 835, 842 (9th Cir. 2004).

Plaintiffs' UCL claim is dismissed with leave to amend to address the many issues identified.

E. Declaratory Relief

Finally, each defendant argues that plaintiffs' declaratory relief claim - seeking a declaration that defendants violated various statutes - fails because all of the other claims fail and because the request for declaratory relief is on face duplicative of their other claims. Given the unsettled status of the pleadings and the evolving nature of plaintiffs' theories of liability, the motions to dismiss the declaratory relief claim as duplicative are DENIED without prejudice. However, when plaintiffs amend their complaint, they should identify the scope of declaratory relief sought and in particular provide examples of declaratory relief sought that are not merely duplicative of the scope of their statutory claims.

II. BREACH OF CONTRACT CLAIM

Plaintiffs separately allege a breach of contract claim against DeviantArt based on the agreement of McKernan and unspecified “others” to DeviantArt's website Terms of Service (“TOS,” Compl. Ex. 17, Dkt. No. 1-17) and DeviantArt's “Privacy Statement.” In the Complaint, plaintiffs allege that unspecified provisions in the TOS control how DeviantArt can use artists' intellectual property and private information and claim that DeviantArt violated those provisions. Compl. ¶ 232. Plaintiffs also rely on unspecified TOS provisions prohibiting users from taking content “for any commercial purpose” and prohibiting commercial advertising on DeviantArt's website without DeviantArt's written approval. Compl. ¶ 124. Plaintiffs contend that DeviantArt itself breached these provisions by:

Provisions of the TOS are quoted in the Complaint and the TOS are attached as an exhibit to the Complaint. The TOS, therefore, are incorporated by reference.

a. Sharing Plaintiffs' and the Class's personal data with unauthorized third parties in violation of the DeviantArt Privacy Statement;
b. Selling and distributing Plaintiffs' and the Class's personal data in contravention of the DeviantArt's Policies;
c. Use of Plaintiffs' and the Class's personal data after the DeviantArt Privacy Statement explicitly claims it will be deleted;
d. Use and distribution of Plaintiffs' and the Class's personal data outside the limitations set forth in the DeviantArt Privacy Statement.
Compl. ¶ 232. They also assert that DeviantArt “was aware or reasonably should have been aware” that Stability was “acting in violation of those terms.” Id. ¶ 125.

DeviantArt moves to dismiss the breach claim, contending that plaintiffs have failed to allege acts of DeviantArt's conduct (as opposed to Stability's conduct) that DeviantArt took or failed to take that could constitute a breach of specific provisions of its TOS or privacy statement. It points out that there is no evidence that Stability is bound by DeviantArt's TOS or Privacy Policy that provide the basis of the breach of contract claim, and instead points to provisions in the TOS that give DeviantArt discretion in how to handle infringement on its website. TOS § 3 (“Trademarks”), § 14 (“Password”). It also argues that the only provisions of the TOS relied on by plaintiffs in their Complaint do not restrict DeviantArt but instead restrict users' ability to conduct commercial and other activity on the DeviantArt website. See, e.g., TOS § 19 (“Conduct,” discussing user conduct); § 19A (“Commercial Activities”).

In opposition, plaintiffs focus on § 16 of the TOS, a provision not identified or quoted in their Complaint:

16. Copyright in Your Content

DeviantArt does not claim ownership rights in Your Content. For the sole purpose of enabling us to make your Content available through the Service, you grant to DeviantArt a non-exclusive, royalty-free license to reproduce, distribute, re-format, store, prepare derivative works based on, and publicly display and perform Your Content. Please note that when you upload Content, third parties will be able to copy, distribute and display your Content using readily available tools on their computers for this purpose although other than by linking to your Content on DeviantArt any use by a third party of your Content could violate paragraph 4 of these Terms and Conditions [preserving copyright rights in the original owner of the copyright and disclaiming any ownership interest of DeviantArt in the posted work] unless the third party receives permission from you by license.

Plaintiffs argue that DeviantArt breached this provision when “it incorporated Stable Diffusion into its own AI Image product knowing that Stability had scraped DeviantArt's artists' work.” Oppo to DeviantArt MTD [Dkt. No. 65] at 22. However, section 16 provides a limited license to DeviantArt and warns that third parties may be able to copy and violate content-owners' rights. It does not clearly cover the conduct that plaintiffs accuse DeviantArt of in this suit; offering for use a product that a third party may have created in part by using material posted on DeviantArt's own site. There are no facts alleged supporting an allegation that DeviantArt itself exceeded the scope of the limited license.

DeviantArt also challenges the ability of plaintiff McKernan and the unspecified “others” to sue DeviantArt for breach claims based on contractual provisions prohibiting other users (presumably here, Stability) from using DeviantArt content for commercial uses. The Complaint does not allege and is devoid of facts supporting the inference that Stability is bound by the TOS or that plaintiff McKernan or others are third party beneficiaries of specific provisions in the TOS who may sue to enforce terms of agreements entered between DeviantArt and Stability.

The breach of contract claim is DISMISSED with leave to amend. If plaintiffs attempt to amend this claim, they must identify the exact provisions in the TOS they contend DeviantArt breached and facts in support of breach of each identified provision. To the extent plaintiffs rely on provisions that appear to protect or benefit DeviantArt but not the users, or contracts DeviantArt entered into with other entities, plaintiffs must identify those precise provisions and facts in support of breach, but also facts supporting plaintiffs' theory that they are intended third party beneficiaries of those provisions.

III. CLASS ALLEGATIONS

Midjourney separately moves to strike plaintiffs' class action allegations. Dkt. No. 523 at 21-25. Midjourney recognizes my many decisions rejecting pleading-stage challenges to whether allegations support class certification under Rule 23. Id. at 22. It nonetheless argues that the class allegations should be stricken in this case because no damages class or injunctive relief class can be pleaded given the inherently fact-intensive inquiries regarding copyright ownership and registration, similarity and confusion determinations, as well as standing and consent to use issues that routinely arise in copyright and other branches of intellectual property litigation.

The motion to strike the class allegations is DENIED. Whether or not the types of claims that remain after the pleadings are settled are certifiable - in whole or in part, for damages under Rule 23(b)(3) or for resolution of common issues under Rule 23(b)(1) or (b)(2) - is better determined at the class certification stage and not at the motion to dismiss stage. At this juncture, precluding the possibility of resolution of issues or claims through a class action is premature.

IV. DEVIANTART SPECIAL MOTION TO STRIKE

DeviantArt moves to strike the plaintiffs' right of publicity claims, arguing their conduct is protected activity under California's anti-SLAPP statute. California Code of Civil Procedure § 425.16. Assuming the anti-SLAPP statute applies - in other words, that DeviantArt's conduct in providing the DreamUp program is expressive conduct protected by the statute - the merits of the special motion to strike depends on the same arguments defendants assert in their motions to dismiss, namely, whether the right to publicity claims are preempted by the Copyright Act and whether they have otherwise been adequately alleged.

The other defendants joint DeviantArt's special motion to strike. Dkt. Nos. 54, 59, 79, 80.

Because I have dismissed the right to publicity claim with leave to amend, I defer ruling on DeviantArt's special motion to strike. When plaintiffs reallege the right to publicity claim, then DeviantArt may renew its special motion to strike, and I will consider that motion on the merits at that juncture.

CONCLUSION

The motions to dismiss are GRANTED in full, except for the direct copyright infringement claim asserted by plaintiff Anderson against Stability. Plaintiffs are given leave to amend and attempt to cure the deficiencies identified above. The amended complaint, if any, must be filed within thirty (30) days of the date of this Order.

IT IS SO ORDERED.


Summaries of

Andersen v. Stability AI Ltd.

United States District Court, Northern District of California
Oct 30, 2023
23-cv-00201-WHO (N.D. Cal. Oct. 30, 2023)

granting motion to dismiss with leave to amend

Summary of this case from Lawrence v. Finicity Corp.
Case details for

Andersen v. Stability AI Ltd.

Case Details

Full title:SARAH ANDERSEN, et al., Plaintiffs, v. STABILITY AI LTD., et al.…

Court:United States District Court, Northern District of California

Date published: Oct 30, 2023

Citations

23-cv-00201-WHO (N.D. Cal. Oct. 30, 2023)

Citing Cases

Tremblay v. Openai, Inc.

In the Ninth Circuit, a plaintiff “must make an affirmative showing, such as by demonstrating a past ‘pattern…

Lawrence v. Finicity Corp.

In this regard, Plaintiff's complaints are not too different from the artists complaining about computer…