Ziff Davis et al. v. OpenAI et al.

• Court: District Court, S.D. New York
• Decided: December 15, 2025
• Docket: 1:25-cv-04315
• Status: Unknown

Opinion

UNITED STATES DISTRICT COURT SOUTHERN DISTRICT OF NEW YORK

IN RE:

25-md-3143 (SHS) (OTW) OPENAI, INC. COPYRIGHT INFRINGEMENT LITIGATION OPINION & ORDER

This Document Relates To: Ziff Davis et al. v. OpenAI et al., 25-cv-4315 SIDNEY H. STEIN, U.S. District Judge. OpenAI1 has moved to dismiss certain claims in Ziff Davis’s2 First Amended 0F 1F Complaint and to stay this action in part with respect to certain causes of action and models. (Dkt. Nos. 405 (Mot. to Dismiss), 134 (Mot. to Stay).) OpenAI’s motion to dismiss is granted with respect to claims 4 and 5, is granted in part and denied in part with respect to claim 8, and is denied with respect to claims 3, 6, and 7. OpenAI’s motion to stay is denied as to claim 7 and is moot as to claims 4 and 5 in light of the Court’s dismissal of those claims. OpenAI’s motion to stay is granted as to the models not already in this multidistrict litigation (“MDL”), namely, the o1, o1 mini, o1-pro, GPT-4.1, GPT-4.5, o3, o3-mini, o4-mini, and GPT-5 models. I. BACKGROUND All factual allegations in the First Amended Complaint (“FAC”) are assumed to be true for the purposes of OpenAI’s motion to dismiss. Faber v. Metro. Life Ins. Co., 648 F.3d 98, 104 (2d Cir. 2011). Ziff Davis has been a publisher of journalism for nearly 100 years and currently owns more than 45 digital media publications and internet brands, including IGN, Mashable, CNET, ZDNET, PCMag, Lifehacker, BabyCenter, and Everyday Health, which

1 Defendants OpenAI, Inc., OpenAI GP LLC, OpenAI, LLC, OpenAI OpCo LLC, OpenAI Global LLC, OAI Corporation, LLC, OpenAI Holdings, LLC, OpenAI Startup Fund I LP, OpenAI Startup Fund GP I LLC, and OpenAI Startup Fund Management LLC are referred to collectively as “OpenAI.” 2 Plaintiffs Ziff Davis, Inc., Ziff Davis, LLC, IGN Entertainment, Inc., Everyday Health Media, LLC, Mashable, Inc., and CNET Media, Inc. are referred to collectively as “Ziff Davis.” collectively produce nearly 2 million articles supported by rigorous research, reporting, and product testing annually. (Dkt. No. 300 (FAC) ¶¶ 2, 33, 39.) Ziff Davis’s portfolio focuses on content in “discrete and consumer-rich media ‘vertical’ categories” including Technology and Shopping, Gaming and Entertainment, and Health and Wellness. (Id. ¶ 35.) Ziff Davis’s digital media publications receive hundreds of millions of unique user visits per month and have won numerous publishing awards. (Id. ¶¶35–36.) Ziff Davis does not typically place its media content behind “paywalls” and instead generates revenue largely through advertising, content licensing, syndication, and commissions on products that users purchase via links in Ziff Davis’s product reviews. (Id. ¶¶ 47–48, 52, 196.) Ziff Davis owns federally registered trademarks associated with many of its brands, seven of which—MASHABLE, LIFEHACKER, CNET, ZDNET, PCMAG, BABYCENTER, and IGN—Ziff Davis asserts are famous (the “Ziff Davis Marks”). (Id. ¶¶ 55–57.) OpenAI is a technology company founded in December 2015 as a non-profit organization focused on artificial intelligence (“AI”) research and relaunched in March 2019 as a for-profit enterprise. (Id. ¶¶ 58, 61.) OpenAI develops large language models (“LLMs”), including its “Generative Pretrained Transformer” or “GPT” series of LLMs. (Id. ¶ 3.) When developing these LLMs, OpenAI compiles large datasets comprised of human-authored text such as articles, books, and scraped web data, copies these datasets into its storage systems, then uses these datasets to “train” its LLMs. (Id. ¶¶ 76– 79.) To train an LLM, OpenAI breaks the text of the gathered third-party content into “tokens,” which consist of words or portions of words. (Id. ¶ 79.) During training, some token are “masked” and the LLM attempts to predict what the masked token is. (Id. ¶ 80.) Over time, this process “trains” the LLM to predict more accurately the correct masked token, which ultimately allows the LLM to produce text resembling human- authored text. (Id. ¶ 80–83.) One consequence of the training process is that LLMs exhibit a behavior known as “memorization” where, when provided with the right prompt, an LLM can reproduce verbatim substantial portions of the materials that were used to train the LLM. (Id. ¶¶ 87–88.) OpenAI has created lucrative products using its LLMs, including its chatbots ChatGPT, ChatGPT Search, and ChatGPT Deep Research, and licenses its products to businesses that build their own offerings using OpenAI’s LLMs. (Id. ¶¶ 92, 96.) ChatGPT in particular has grown immensely in popularity since its launch, with 400 million weekly active users as of February 2025 and 800 million weekly active users by early April 2025. (Id. ¶ 94.) OpenAI used copyrighted works created and owned by Ziff Davis during its development and operation of its LLMs. (Id. ¶ 107.) OpenAI copied Ziff Davis’s works into its LLM training datasets using existing pools of scraped website content and by directly scaping content from Ziff Davis’s websites. (Id. ¶ 108.) These copies did not include copyright management information and were created by ChatGPT’s web- scraping tool, “GPTBot,” despite Ziff Davis’s inclusion of robots.txt files in its websites’ code, which allegedly instructed GPTBot not to scrape Ziff Davis’s websites. (Id. ¶¶ 109, 116–28.) The outputs of OpenAI’s LLMs sometimes include content drawn or copied from Ziff Davis’s copyrighted works, and OpenAI uses those outputs to further train its LLMs. (Id. ¶¶ 111–12.) Some outputs of OpenAI’s LLMs misleadingly or inaccurately attribute content to Ziff Davis. (Id. ¶¶ 174–77.) The First Amended Complaint asserts nine causes of action: (1) copyright infringement by OpenAI based on its use of Ziff Davis’s works to train OpenAI’s LLMs, (2) copyright infringement based on outputs of OpenAI’s LLMs, (3) contributory copyright infringement, (4) common law unjust enrichment, (5) circumvention of technological measures in violation of 17 U.S.C. § 1201(a)(1), (6) removal of copyright management information in violation of 17 U.S.C. § 1202(b)(1), (7) distribution of works with copyright management information removed in violation of 17 U.S.C. § 1202(b)(3), (8) trademark dilution, and (9) dilution and injury to business reputation in violation of Delaware state law. II. MOTION TO DISMISS OpenAI has moved to dismiss claims 3, 4, 5, 6, 7, and 8. The Court grants OpenAI’s motion to dismiss claims 4 and 5, grants in part and denies in part the motion to dismiss claim 8, and denies the motion to dismiss claims 3, 6, and 7. A. Motion to Dismiss Standard “To survive a motion to dismiss, a complaint must contain sufficient factual matter, accepted as true, to ‘state a claim to relief that is plausible on its face.’” Ashcroft v. Iqbal, 556 U.S. 662, 678 (2009) (quoting Bell Atl. Corp. v. Twombly, 550 U.S. 544, 570 (2007)). When considering a Rule 12(b)(6) motion to dismiss, a court must “draw all reasonable inferences in Plaintiffs’ favor, assume all well-pleaded factual allegations to be true, and determine whether they plausibly give rise to an entitlement to relief.” Faber, 648 F.3d at 104 (internal quotations marks and citation omitted). In so doing, the Court may consider the facts stated on the face of the complaint as well as “any written instrument attached to the complaint as an exhibit, any statements or documents incorporated in it by reference, and any document upon which the complaint heavily relies.” In re Thelen LLP, 736 F.3d 213, 219 (2d Cir. 2013). Although the Court must accept as true all factual allegations in the complaint, “[t]hreadbare recitals of the elements of a cause of action, supported by mere conclusory statements, do not suffice.” Iqbal, 556 U.S. at 678. By the same token, “[t]he choice between two plausible inferences that may be drawn from factual allegations is not a choice to be made by the court on a Rule 12(b)(6) motion.” Anderson News, L.L.C. v. Am. Media, Inc., 680 F.3d 162, 185 (2d Cir. 2012). “[F]act-specific question[s] cannot be resolved on the pleadings.” Id. (alterations in original) (quoting Todd v. Exxon Corp., 275 F.3d 191, 203 (2d Cir. 2001)). B. Analysis 1. Claim 4: Common Law Unjust Enrichment Claim 4 alleges that OpenAI has been unjustly enriched by taking Ziff Davis’s copyrighted works and using them to train OpenAI’s models without compensating Ziff Davis for the use of the copyrighted works. In response, OpenAI contends that Ziff Davis’s unjust enrichment claim is preempted by the Copyright Act. a. Relevant Legal Standards “Section 301 of the Copyright Act provides for ‘the preemption of state law claims that are interrelated with copyright claims in certain ways.’” N.Y. Times Co. v. Microsoft Corp., 777 F. Supp. 3d 283, 320 (S.D.N.Y. 2025) (quoting Nat’l Basketball Ass’n v. Motorola, Inc., 105 F.3d 841, 848 (2d Cir. 1997)).3 Section 301(a) lays out a two-part test for the 2F preemption of state laws. A state law claim is preempted when it (1) concerns “works of authorship that . . . come within the subject matter” of the Copyright Act (the “subject matter requirement”) and (2) seeks to vindicate “legal or equitable rights that are equivalent to any of the exclusive rights” protected by Section 106 the Copyright Act (the “general scope requirement”). See 17 U.S.C. § 301(a). b. Application The subject matter requirement is clearly satisfied here because the subject matter of copyright extends to “literary works” such as the Ziff Davis works at issue in this action. 17 U.S.C. § 102(a)(1). Ziff Davis does not challenge the subject matter prong on this motion. (Dkt. No. 494 (“MTD Opp.”) at 11.) The general scope requirement necessitates somewhat greater examination.

3 Section 301 states that “all legal or equitable rights that are equivalent to any of the exclusive rights within the general scope of copyright . . . in works of authorship that are fixed in a tangible medium of expression and come within the subject matter of copyright . . . are governed exclusively by this title.” 17 U.S.C. § 301(a). Where a state law claim includes additional or different elements to the analogous claim under the federal Copyright Act, the similar state law claim is not preempted, even if it would otherwise apparently fall within the general scope of federal copyright law. See Nat’l Basketball Ass’n, 105 F.3d at 850–53 (determining that a state law misappropriation claim contained “extra elements” and was therefore not preempted by section 301 of the Copyright Act). OpenAI, citing the U.S. Court of Appeals for the Second Circuit’s decision in Briarpatch Ltd. v. Phoenix Pictures, Inc., 373 F.3d 296 (2d Cir. 2004), contends that Ziff Davis’s unjust enrichment claim contains no “extra elements” and is therefore preempted by the Copyright Act. (See Dkt. No. 407 (“MTD”) at 15–16.) Ziff Davis urges that its unjust enrichment claim is not preempted because “unjust enrichment claims that include allegations of certain types of improper ‘use’ (particularly to develop new technology) can provide the types of ‘extra elements’ that can defeat a preemption defense.” (MTD Opp. at 12.) “Courts have generally concluded that the theory of unjust enrichment protects rights that are essentially ‘equivalent’ to rights protected by the Copyright Act; thus, unjust enrichment claims related to the use of copyrighted material are generally preempted.” Ardis Health, LLC v. Nankivell, No. 11-cv-5013, 2012 WL 5290326, at *10 (S.D.N.Y. Oct. 23, 2012) (quoting Weber v. Geffen Records, Inc., 63 F. Supp. 2d 458, 462 (S.D.N.Y. 1999)); see, e.g., Briarpatch, 373 F.3d at 306–07 (quoting with approval 1 Nimmer on Copyright § 1.01[B][1][g] (2003) for the proposition that “[a] state law cause of action for unjust enrichment or quasi contract should be regarded as an ‘equivalent right’ and hence, pre-empted insofar as it applies to copyright subject matter”); Einiger v. Citigroup, Inc., No. 14-cv-4570, 2014 WL 4494139, at *7 (S.D.N.Y. Sep. 12, 2014). The Court discerns no reason to depart from that uniform conclusion. Ziff Davis’s attempt to save its claim from preemption by characterizing it as based on “use” rather than “reproduction” is unpersuasive. The basis of Ziff Davis’s unjust enrichment claim is that OpenAI used Ziff Davis’s copyrighted works in an unauthorized way. The fundamental nature of this claim is equivalent to a claim to enforce exclusive rights within the general scope of copyright. Cf. Doe 1 v. GitHub, Inc., No. 22-cv-6823, 2024 WL 235217, at *7–8 (N.D. Cal. Jan. 22, 2024) (dismissing on preemption grounds claims, including unjust enrichment claims, “principally concern[ing] the unauthorized reproduction of” the plaintiff’s copyrighted works in training of generative AI because the claims “fall under the purview of the Copyright Act”); Anderson v. Stability AI, 744 F. Supp. 3d 956, 972–73 (N.D. Cal. 2024) (dismissing as preempted claims based on the use of the plaintiffs’ copyrighted works in the training of a generative AI product). Ziff Davis’s unjust enrichment claim is therefore dismissed as preempted by the Copyright Act. 2. Claim 5: Circumvention of Technological Measures Claim 5 alleges that Ziff Davis implemented a technological measure—namely, robots.txt files—designed to control access to its copyrighted works but OpenAI improperly circumvented that technological measure to gain access to Ziff Davis’s works. a. Relevant Legal Standards Section 1201(a) of the Digital Millenium Copyright Act (“DMCA”) provides that “[n]o person shall circumvent a technological measure that effectively controls access to a work protected under [the Copyright Act].” 17 U.S.C. § 1201(a)(1)(A). “[T]o ‘circumvent a technological measure’ means to descramble a scrambled work, to decrypt an encrypted work, or otherwise to avoid, bypass, remove, deactivate, or impair a technological measure, without the authority of the copyright owner.” 17 U.S.C. § 1201(a)(3)(A). “[A] technological measure ‘effectively controls access to a work’ if the measure, in the ordinary course of its operation, requires the application of information, or a process or a treatment, with the authority of the copyright owner, to gain access to the work.” Id. § 1201(a)(3)(B). To prevail on its section 1201(a) claim, therefore, Ziff Davis must allege both that the robots.txt files are a “technological measure that effectively controls access” to a copyrighted work and that OpenAI “circumvented” that technological measure. b. Application The FAC fails to allege that robots.txt files are a “technological measure that effectively controls access” to copyrighted works or that OpenAI “circumvented” that technological measure. First, Ziff Davis fails to allege that robots.txt instructions are a technological measure that effectively controls access to Ziff Davis’s copyrighted works. According to the FAC, “robots.txt directives . . . are machine-readable instructions . . . which tell web crawlers which areas of the site the bot is allowed or disallowed from accessing and indexing” and, “[i]n order to bypass a robots.txt ‘disallow’ directive, a scraper must actively and intentionally override these explicit technical directives to access the protected content.” (FAC ¶¶ 117–18.) The FAC further alleges that OpenAI “actively ignore[d]” the robots.txt directive that OpenAI’s web crawlers not scrape Ziff Davis’s web pages. (Id. ¶ 121.) These allegations do not establish that robots.txt files are a “measure designed to thwart unauthorized access” to Ziff Davis’s protected works. LivePerson, Inc. v. 24/7 Customer, Inc., 83 F. Supp. 3d 501, 510–11 (S.D.N.Y. 2015). Robots.txt files instructing web crawlers to refrain from scraping certain content do not “effectively control” access to that content any more than a sign requesting that visitors “keep off the grass” effectively controls access to a lawn. On Ziff Davis’s own telling, robots.txt directives are merely requests and do not effectively control access to copyrighted works. A web crawler need not “appl[y] . . . information, or a process or a treatment,” in order to gain access to web content on pages that include robots.txt directives, 17 U.S.C. § 1201(a)(3)(B); it may access the content without taking any affirmative step other than impertinently disregarding the request embodied in the robots.txt files. The FAC therefore fails to allege that robots.txt files are a “technological measure that effectively controls access” to Ziff Davis’s copyrighted works, and the DMCA section 1201(a) claim fails for this reason.4 3F Second, even if robots.txt files were a “technological measure that effectively controls access” to Ziff Davis’s works, the FAC fails to allege that OpenAI “circumvented” the robots.txt files. A plaintiff asserting a DMCA section 1201 claim must allege that a defendant “affirmatively perform[ed] an action that disables or voids” the technological control measure, akin to descrambling or decrypting a work or “breaking and entering (or hacking) into computer systems.” LivePerson, 83 F. Supp. 3d at 509. At most, Ziff Davis alleges that OpenAI disregarded the instructions that were contained in robots.txt files. This is not “circumvention” under the DMCA, and Ziff Davis’s claim fails for this reason as well. 3. Claim 7: Distribution of Works with Copyright Management Information Removed Claim 7 alleges that OpenAI distributed copies of Ziff Davis’s copyrighted works with the copyright management information (“CMI”) removed in violation of section 1202(b)(3) of the DMCA.

4 Ziff Davis relies on Healthcare Advocates, Inc. v. Harding, Early, Follmer & Frailey, 497 F. Supp. 2d 627 (E.D. Pa. 2007), to rebut the conclusion that robots.txt files are not technological measures that effectively control access to Ziff Davis’s copyrighted works. Healthcare Advocates involved a unique factual circumstance in which, unlike in this case, a user would actually be prevented by a robots.txt file from viewing an archived website using the “Wayback Machine.” Id. at 643. Further, the court in Healthcare Advocates went out of its way to emphasize that its finding that the robots.txt files in that case were “technological measure[s]” covered by the DMCA “should not be interpreted as a finding that a robots.txt file universally qualifies as a technological measure that controls access to copyrighted works under the DMCA.” Id. a. Relevant Legal Standards The DMCA provides at section 1202(b)(3) that “[n]o person shall, without the authority of the copyright owner or the law . . . distribute . . . copies of works . . . knowing that copyright management information has been removed or altered without authority of the copyright owner or the law . . . knowing, or . . . having reasonable grounds to know, that it will induce, enable, facilitate, or conceal an infringement of any right under [the Copyright Act].” To plead a claim under section 1202(b)(3), a plaintiff must allege “(1) the existence of CMI in connection with a copyrighted work; and (2) that a defendant distributed works or copies of work; (3) while knowing that CMI has been removed or altered without authority of the copyright owner or the law; and (4) while knowing, or having reasonable grounds to know that such distribution will induce, enable, facilitate, or conceal an infringement.” Mango v. Buzzfeed, Inc., 970 F.3d 167, 171 (2d Cir. 2020) (cleaned up). b. Application OpenAI contends that Ziff Davis’s section 1202(b)(3) claim fails because Ziff Davis has not alleged distribution of works lacking CMI and because Ziff Davis does not allege that OpenAI distributed complete copies of Ziff Davis’s works without CMI. The Court finds that the FAC states a claim under section 1202(b)(3) because Ziff Davis alleges both distribution and that OpenAI distributed complete copies of Ziff Davis’s works without CMI. i. Distribution OpenAI contends that the FAC fails to allege distribution for two reasons: (1) the FAC alleges only that Ziff Davis’s own lawyers or experts have caused ChatGPT to generate outputs that are copies of Ziff Davis’s works without CMI, not that such copies have been distributed to third parties, and (2) ChatGPT’s outputs constitute “public displays” rather than “distribution” of copies of Ziff Davis’s works without CMI. (MTD at 18–20.) OpenAI’s first contention misses the mark because the FAC alleges that “OpenAI reproduced and created derivative, and in some instances identical, versions of [Ziff Davis’s] copyrighted works from which CMI was intentionally removed, and then distributed those versions to third parties without permission and with knowledge that the CMI had been removed.” (FAC ¶ 258.) Far from being “conclusory,” this allegation is supported by Ziff Davis’s further allegations that it was able to prompt ChatGPT to create and provide such outputs to Ziff Davis itself. (Id. ¶¶166–68.) This is sufficient to render plausible the FAC’s allegation that such outputs have also been generated for third parties. This is all that is required at the pleading stage, and the Court will not dismiss the claim on the basis of implausibility when the FAC’s allegations establish ChatGPT’s capacity for generating outputs of the type Ziff Davis complains of. As to OpenAI’s second contention, the FAC’s allegations are sufficient at the pleading stage to establish that ChatGPT’s outputs containing copies of Ziff Davis’s copyrighted works with CMI removed go beyond mere public display and constitute distribution. OpenAI is correct that courts have understood “distribution” under the DMCA to require a “sale or transfer of ownership extending beyond that of a mere public display” of a copy of a copyrighted work. Wright v. Miah, No. 22-cv-4132, 2023 WL 6219435, at *7 (E.D.N.Y. Sep. 7, 2023). But the FAC alleges more than mere public display. It alleges that OpenAI’s LLMs create and provide to third parties copies of Ziff Davis’s works without CMI and that OpenAI “then purports to transfer ownership of the infringing output . . . to its commercial and individual users” pursuant to OpenAI’s Terms of Use. (FAC ¶ 170.) The FAC also alleges that “hundreds of full copies of the body text of Ziff Davis Works” appear in the sample of the so-called WebText training data that OpenAI made publicly available on the website GitHub (id. ¶¶ 130, 138–40)5 4F and that OpenAI published this data under a license that granted “any person obtaining a copy of” the data the rights to “use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies” of the data (see FAC ¶ 140 n.66; MTD Opp. at 17 (quoting the terms of the license appearing at https://github.com/openai/gpt-2-output-dataset)).6 5F In response, OpenAI asserts without citation that “a contractual provision cannot transform a display into a distribution.” (MTD Reply at 9 n.4.) This ignores the importance of a “sale or transfer of ownership”—which, needless to say, would often be

5 In its reply memorandum, OpenAI raised for the first time the contention that claims based on the WebText data are time-barred. (See Dkt. No. 565 (“MTD Reply”) at 10.) It is well established that a party may not raise arguments for the first time in its reply brief; the Court will not consider this argument. See, e.g., McBride v. BIC Consumer Prods. Mfg. Co., Inc., 583 F.3d 92, 96 (2d Cir. 2009) (“[W]e ordinarily will not consider issues raised for the first time in a reply brief.”). 6 The Court may consider the terms of the license appearing on the web page on which the WebText data was published for the purposes of this motion because the FAC makes “clear, definite and substantial reference” to that web page. See Videri, Inc. v. ONAWHIM (OAW) Inc., No. 23-cv-2535, 2024 WL 4027980, at *7 (S.D.N.Y. Sep. 3, 2024) (considering the content of a web page for the purpose of a motion to dismiss because the complaint made a “clear, definite and substantial reference” to the page, including by linking to the page); see also In re OpenAI, Inc. Copyright Infringement Litig., No. 25-md-3143, 2025 WL 3003339, at *6 (S.D.N.Y. Oct. 27, 2025) (“To be incorporated by reference, the complaint must make a clear, definite and substantial reference to the documents.” (quoting Lateral Recovery, LLC v. Cap. Merchant Servs., LLC, 632 F. Supp. 3d 402, 436 (S.D.N.Y. 2022))). accomplished by contract (such as OpenAI’s Terms of Use or the license accompanying the WebText data)—in distinguishing between a mere public display and distribution. Taken as true—as they must be on this motion to dismiss—the FAC’s allegations establish that the outputs of OpenAI’s LLMs went beyond public display and were “distributions” of copies of Ziff Davis’s copyrighted works with CMI removed. ii. Complete Copies Second, the FAC alleges that OpenAI distributed complete copies of Ziff Davis’s works with CMI removed, not only that OpenAI distributed partial copies or excerpts. The FAC alleges that “Ziff Davis has . . . documented generation of verbatim copies, close paraphrases, and substantiated derivatives of Ziff Davis Works . . . in OpenAI output.” (FAC ¶ 166.) Further, Exhibits D and E to the FAC purport to contain several complete copies of Ziff Davis works included in ChatGPT outputs and in the WebText data published by OpenAI without CMI. (E.g., FAC, Ex. D; Ex. E at 1–3.) These examples are sufficient to establish at the pleading stage that Ziff Davis has made out a claim that OpenAI distributed complete copies of Ziff Davis’s works with CMI removed.7 6F 4. Claim 8: Trademark Dilution Claim 8 alleges that the Ziff Davis Marks are famous and that OpenAI has impermissibly diluted those marks by providing the famous marks in the LLM outputs in violation of 15 U.S.C. § 1125(c). a. Relevant Legal Standards 15 U.S.C. § 1125(c) provides a cause of action to “the owner of a famous mark that is distinctive” when another, “at any time after the owner’s mark has become famous, commences use of a mark or trade name in commerce that is likely to cause dilution by blurring or dilution by tarnishment of the famous mark.” To state a claim for trademark dilution under section 1125(c), a plaintiff must allege “(1) the mark is famous; (2) [the] defendant’s use of the mark is made in commerce; (3) the defendant used the mark after

7 At oral argument on this motion, OpenAI contended that some of the ChatGPT outputs in Exhibit E to the FAC in fact do contain CMI because they include links to Ziff Davis’s works and because a link to CMI itself qualifies as CMI under the DMCA. See 17 U.S.C. § 1202(c)(7) (defining CMI to mean “identifying numbers or symbols referring to [CMI] or links to such information”). The Court need not determine if the examples OpenAI highlights from Exhibit E to the FAC in fact contain CMI because the complete copies of Ziff Davis’s works in the published WebText data do not contain CMI or links to CMI. (See FAC, Ex. D.) the mark is famous; and (4) the defendant’s use of the mark is likely to dilute the quality of the mark by blurring or tarnishment.” DigitAlb, Sh.a v. Setplex, LLC, 284 F. Supp. 3d 547, 557 (S.D.N.Y. 2018). Fame is the “key ingredient” in a federal trademark dilution claim, Savin Corp. v. Savin Grp., 391 F.3d 439, 449 (2d Cir. 2004), and “[t]he requirement that the mark be ‘famous’ and ‘distinctive’ significantly limits the pool of marks that may receive dilution protection,” Starbucks Corp. v. Wolfe’s Borough Coffee, Inc., 588 F.3d 97, 105 (2d Cir. 2009). “[A] mark is famous if it is widely recognized by the general consuming public of the United States as a designation of source of the goods or services of the mark’s owner.” 15 U.S.C. § 1125(c)(2)(A). In determining whether a mark is “famous” for the purposes of a trademark dilution claim, a court “may consider all relevant factors,” including “[t]he duration, extent, and geographic reach of advertising and publicity of the mark, whether advertised or publicized by the owner or third parties”; “[t]he amount, volume, and geographic extent of sales of goods or services offered under the mark”; “[t]he extent of actual recognition of the mark”; and “[w]hether the mark was registered under the Act of March 3, 1881, or the Act of February 20, 1905, or on the principal register.” Id. A complaint plausibly alleges that a trademark is famous when the allegations include attributes of the mark such as “nationwide recognition and respect,” e.g., Lewittes v. Cohen, No. 03-cv-189, 2004 WL 1171261, at *6 (S.D.N.Y. May 26, 2004); continuous and pervasive use of the mark, e.g., A.V.E.L.A., Inc. v. Estate of Marilyn Monroe, LLC, 131 F. Supp. 3d 196, 216 (S.D.N.Y. 2015); substantial investments in promoting and advertising the mark throughout the United States and internationally, e.g., New York City Triathlon, LLC v. NYC Triathlon Club, Inc., 704 F. Supp. 2d 305, 321–22 (S.D.N.Y. 2010); “significant publicity” relating to the marks, e.g., A.V.E.L.A., 131 F. Supp. 3d at 216; and that “products bearing the [plaintiff’s] [m]arks are sold throughout the United States,” id. A complaint does not sufficiently allege a mark’s fame when it includes only “spare, conclusory allegations” that the trademark holder has “expended substantial time, effort, money, and resources advertising and promot[ing]” a trademarked product, CDC Newburgh Inc. v. STM Bags, LLC, 692 F. Supp. 3d 205, 235 (S.D.N.Y. 2023), or when the complaint includes only a single, conclusory allegation that a trademark is “widely recognized by the general public,” DigitAlb, 284 F. Supp. 3d at 558. b. Application OpenAI first contends that Ziff Davis has failed to establish that the Ziff Davis Marks are “famous” because the FAC alleges only that Ziff Davis’s brands are famous, not that the corresponding marks are themselves famous. (MTD at 22–23.) OpenAI also contends that Ziff Davis failed to allege widespread or general recognition of its marks and instead alleged only niche fame. (Id. at 23–24.) The Court finds that the FAC sufficiently alleges that the MASHABLE mark, as opposed to the Mashable brand more generally, is “famous.” Trademark dilution claims concern marks, not brands or publications generally. See 15 U.S.C. § 1125(c)(2)(A) (defining fame in reference to “a mark”). The FAC alleges that the Mashable publication “was founded in 2005,” was named “one of the 25 best blogs in the world” by Time Magazine by 2009, “had a reported average of 12.9 million monthly unique visitors” over the last twelve months of available data, had 22.7 million followers across various social media platforms, “attracts readers from all over the world,” and ”is considered a leading source for technology and entertainment news worldwide, winning various awards, and generating a substantial amount of the total 2024 revenue.” (FAC ¶ 56.) And, crucially, the FAC alleges that all of this happens “under the MASHABLE Mark,” which has “achieved widespread recognition and fame by virtue of [its] respective scope of use, duration of use in interstate commerce, and the geographic reach of advertising and publicity related to” the Mashable brand. (Id.) These allegations sufficiently establish for the purposes of OpenAI’s motion to dismiss that the MASHABLE mark “is widely recognized by the general consuming public of the United States as a designation of source of” articles from the Mashable publication. Cf. N.Y. Times Co., 777 F. Supp. 3d at 324 (determining that complaint sufficiently alleged marks’ fame based on allegations that publications operating under marks existed for significant period of time, several marks were registered, publications circulated throughout United States, publications’ articles received significant attention and praise from other news outlets, millions of consumers accessed publications, and publications won many awards). The FAC does not, however, sufficiently allege that any of the other Ziff Davis Marks—as opposed to the brands associated with those marks—is “famous.” Unlike the allegations regarding the Mashable publication’s operation under the MASHABLE mark, the FAC does not allege that the Lifehacker, CNET, ZDNET, PCMag, BabyCenter, or IGN publications operate under the LIFEHACKER, CNET, ZDNET, PCMAG, BABYCENTER, or IGN marks. (See FAC ¶ 56.) Though Ziff Davis contends in its opposition to the motion to dismiss that “the public knows, refers to, and recognizes each publication by its corresponding” Ziff Davis Mark (MTD Opp. at 23), nothing in the FAC alleges that these brands in fact operate under widely recognized marks. Instead, the FAC alleges only facts arguably suggesting that the Lifehacker, CNET, ZDNET, PCMag, BabyCenter, and IGN brands are well known, including that the websites for these six brands averaged between 3.2 million and 23.9 million unique visitors each month and had between 1.1 million and 53.5 million followers on social media platforms as of June 2025. (FAC ¶ 56.) Ziff Davis does not allege the extent of promotion or advertising associated with any of the marks. Ziff Davis also fails to allege revenue attributable to any marks and the associated publications individually, instead providing broad revenue figures for Ziff Davis’s “Technology & Shopping category,” “Health & Wellness category,” and “Game & Entertainment category.” (Id.) Cf. Luv N’ Care, Ltd. v. Regent Baby Prods. Corp., 841 F. Supp. 2d 753, 757 (S.D.N.Y. 2012) (determining plaintiffs failed to allege marks were famous when complaint alleged revenue of all goods sold under plaintiffs’ trademarks rather than revenue of goods sold under specific marks at issue). The Court cannot conclude based on the FAC’s allegations that the marks associated with the Lifehacker, CNET, ZDNET, PCMag, BabyCenter, and IGN brands are themselves widely recognizable by the general public, and therefore the Court determines that the FAC fails to establish that these marks are famous for the purposes of Ziff Davis’s trademark dilution claim. 5. Claims 3 and 6: Contributory Infringement and DMCA Section 1202(b)(1) Claim 3 of the FAC alleges that OpenAI has contributed to the infringement by end- users of OpenAI’s LLM-based products and claim 6 alleges that OpenAI intentionally removed CMI when copying Ziff Davis’s works to build its training datasets in violation of 17 U.S.C. § 1202(b)(1). The Court denies OpenAI’s motion to dismiss those claims for the reasons set forth in this Court’s Opinion in New York Times Co. See 777 F. Supp. 3d at 305–16. In sum, Ziff Davis has stated a claim for contributory infringement because “knowledge of specific infringements is not required to support a finding of contributory infringement” and, by alleging “widely publicized instances of copyright infringement” and “numerous examples of infringing outputs in” the FAC, Ziff Davis plausibly alleged end-user infringement and that OpenAI possessed actual or constructive knowledge of this third-party infringement. Id. at 306–08 (citations omitted). Ziff Davis states a claim under section 1202(b)(1) of the DMCA because the FAC alleges a harm that bears a close relationship to traditional copyright infringement, which is sufficient to satisfy the injury-in-fact requirement of Article III of the U.S. Constitution, and because the FAC alleges that OpenAI removed CMI from copies of Ziff Davis’s works with actual or constructive knowledge that such removal could facilitate end-user infringement. Id. at 310–16. III. MOTION TO STAY OpenAI has moved to stay this action with respect to claims 4, 5, and 7 of the FAC, to the extent those claims are not dismissed, and with respect to “any models referenced in the Ziff Davis Complaint that are not already within the MDL’s substantive scope”— that is, with respect to the o1, o1 mini, o1-pro, GPT-4.1, GPT-4.5, o3, o3-mini, o4-mini, and GPT-5 models. (Dkt. No. 136 at 1.) The motion to stay claims 4 and 5 is dismissed as moot in light of the Court’s dismissal of those claims and the motion to stay claim 7 is denied. The motion to stay is granted as to the o1, o1 mini, o1-pro, GPT-4.1, GPT-4.5, o3, o3-mini, o4-mini, and GPT-5 models. A. Standard for Staying Proceedings A district court has inherent power to stay proceedings. Landis v. N. Am. Co., 299 U.S. 248, 254 (1936) (“[T]he power to stay proceedings is incidental to the power inherent in every court to control the disposition of the causes on its docket with economy of time and effort for itself, for counsel, and for litigants.”). In the Second Circuit, courts consider five factors, known as the Kappel factors, in evaluating whether a stay is appropriate: “(1) the private interests of the plaintiffs in proceeding expeditiously with the civil litigation as balanced against the prejudice to the plaintiffs if delayed; (2) the private interest of and burden on the defendants; (3) the interests of the courts; (4) the interests of persons not parties to the civil litigation; and (5) the public interest.” Rankine v. Levi Strauss & Co., 674 F. Supp. 3d 57, 68 (S.D.N.Y. 2023) (citing Kappel v. Comfort, 914 F. Supp. 1056, 1058 (S.D.N.Y. 1996)). B. Analysis 1. Claim 7: Distribution of Works with CMI Removed The Kappel factors do not support staying this action with respect to Ziff Davis’s claim 7, which is based on DMCA section 1202(b)(3) for distribution of works with CMI removed. Ziff Davis (and other plaintiffs in the MDL) assert a claim based on DMCA section 1201(b)(1) for removal of CMI from copies of Ziff Davis’s works, and OpenAI does not seek to stay that claim. The elements of Ziff Davis’s section 1202(b)(3) claim largely overlap with the requirements of its section 1202(b)(1) claim, which is a part of this action and will move forward. The section 1202(b)(3) claim additionally requires Ziff Davis to prove “distribution,” but OpenAI has not persuasively shown that this element would dramatically expand the scope of discovery, unreasonably burden OpenAI, or delay this action and the MDL more generally. Ziff Davis’s and the public’s interest in expeditious resolution of the issues in this MDL counsel against the stay. For these reasons, the Court will not stay proceedings as to claim 7. 2. Additional Models The Kappel factors favor staying this action as to the o1, o1 mini, o1-pro, GPT-4.1, GPT-4.5, o3, o3-mini, o4-mini, and GPT-5 models. The burden on OpenAI of potentially expansive discovery into these models—which are not otherwise at issue in this MDL— would involve additional deponents and documents and would be significant, and the likely delay attendant to such an expansion of the scope of discovery in this MDL at this time would be contrary to the interests of the Court, non-parties, and the public ina speedy resolution of the core copyright issues in this MDL. TV. CONCLUSION OpenAl’s motion to dismiss the First Amended Complaint is granted with respect to claims 4 and 5, is granted in part and denied in part with respect to claim 8, and is denied with respect to claims 3, 6, and 7. OpenAl’s motion to stay is granted as to the models in the FAC that are not already in this MDL, namely, the o1, 01 mini, ol-pro, GPT-4.1, GPT-4.5, 03, 03-mini, o4-mini, and GPT-5 models, but is denied as to claim 7.

Dated: New York, New York December 15, 2025 SO ORDERED:

Sidney FA. Stein, U.S.D,J.