Journalistic Implications of AI: Trademark Dilution in The New York Times v. Microsoft
If someone published thousands of posts every day that incorrectly attributed false or misleading information to a certain newspaper, it would harm the newspaper’s reputation. In most cases, the remedy for this is straightforward: injunctive relief. In 2023, The New York Times found itself in this position, except the source of trademark dilution was slightly more complicated. The publisher allegedly tarnishing their reputation is an AI model—ChatGPT. The case began when a group of newspapers led by The Times sued OpenAI and Microsoft, which owns a 27% stake in OpenAI, for copyright infringement and trademark dilution. [1] While the copyright-centered facets of this case have attracted significant attention, the trademark dilution claims by The Times have received relatively little commentary, despite having monumental implications for the future of journalism and generative AI as a source of news.
The group of trademark holders in the lawsuit, including The New York Times, Chicago Tribune, Mercury News, and Denver Post, is referred to as “the trademark dilution plaintiffs.” To prove trademark dilution, they must demonstrate (i) that they possess a famous mark, which (ii) “has been blurred or tarnished (or both) by someone else’s use of [their] mark or a similar mark.” [2] This blurring and tarnishment occurs when a mark’s distinctiveness or credibility is damaged, respectively. According to section 1125(c) of the Lanham Act, a 1946 law that outlines the registration process, protections, and remedies for trademarks, the bar for a trademark to be considered famous is recognition by the “general consuming public” of the U.S., a burden that OpenAI alleged the trademark dilution plaintiffs failed to meet. [3] However, U.S. District Judge Sidney Stein denied OpenAI’s motion to dismiss the trademark dilution claims, as the trademark dilution plaintiffs successfully argued that the duration, widespread circulation, international acclaim, commercial value, and prestige of their newspapers sufficiently met the Lanham Act standard for famous marks. [4]
The second hurdle for the plaintiffs came in proving the impact of their dilution claims. Prior to 2006, trademark dilution cases followed the precedent set by Moseley v. V Secret Catalogue, Inc. (2003), which relied on the Federal Trademark Dilution Act (FTDA), a law that clarified the requirements for trademark dilution and its remedies. [5] In the case, the plaintiff, Victoria’s Secret sued Moseley, whose store, “Victor’s Little Secret,” sold lower quality clothing, and thus blurred the distinctiveness and tarnished the reputation of V Secret Catalogue. Though the plaintiff provided sufficient evidence to establish likely dilution, they lost the case because they were unable to prove instances of actual harm. The Supreme Court’s ruling in Moseley v. V Secret Catalogue established the precedent that plaintiffs could only receive injunctions “under the FTDA by showing that their trademarks had suffered actual dilution.” [6] However, in 2006, Congress passed the Trademark Dilution Revision Act (TDRA), which relaxed this requirement by lowering plaintiffs’ burden of proof to likely dilution. Thus, the trademark dilution plaintiffs must only prove that OpenAI’s ChatGPT has blurred or tarnished their marks, and that this is not protected under dilution exemptions in the Lanham Act. Additionally, because trademark dilution through tarnishment claims are concerned with the reputation and credibility of marks instead of their distinctiveness, the diluted trademarks do not need to prove competition by OpenAI. They only need to prove that they are likely to experience meaningful harm by dilution. While blurring does require proof of harm to the distinctiveness of a trademark, plaintiffs do not need to prove both blurring and tarnishment to win trademark dilution claims. [7]
By falsely attributing erroneous claims to the diluted trademarks via citations, OpenAI potentially associated its journalism with inaccurate information. The trademark dilution plaintiffs argue that OpenAI’s large language model (LLM) generated outputs that contained hallucinations—fabricated information that AI asserts as truth—and misleading content, which was then incorrectly attributed to newspapers. This association projects an image of poor-quality reporting onto the newspapers, tarnishing the credibility that the diluted trademarks have built over centuries and diluting the prestige and reputation associated with the marks. The Times provided multiple instances of this type of hallucination. For example, The Times “asked ChatGPT twice to identify a quote from an article published in the Washington Post on October 8, 2024, [and] it cited the wrong date and attributed the story to the New York Times without attaching a source.” [8] In other instances, Bing Chat—which uses OpenAI’s LLM—“confidently purported to reproduce” a non-existent paragraph of a times article, suggested a list of heart-healthy foods that “The Times article did not provide,” and “completely fabricated” a study about a “Link between Orange Juice and Non-Hodgkin’s Lymphoma” that The Times never published. [9] Revenue for journalism is largely predicated on the reliability and accuracy of that journalism. This can be severely damaged when OpenAI’s LLM associates the newspapers with misleading coverage. Insofar as OpenAI hallucinates source attributions, there is a clear likelihood of financial harm to the trademark dilution plaintiffs. In other cases, The Times cited instances where ChatGPT generated “near verbatim copies of its articles” with omissions of certain sections that provided critical context to those articles. [10] Litigation about the degree of memorization of newspaper articles as training data primarily relates to the copyright infringement aspect of the case, and is outside the scope of this article; however, ChatGPT’s partial—and, therefore, in the context of journalism, lower-quality—representation of The Times’ reporting further dilutes its trademark by presenting incomplete and flawed explanations of its articles. Since journalistic content that the diluted trademark holders did not publish is attributed to their newspapers, and their content is improperly represented, OpenAI’s LLM output is a clear case of trademark dilution. Even if the content is not false, the fact that it is an improper portrayal of the plaintiffs’ actual journalism creates trademark dilution.
The most salient legal defense for OpenAI is likely the news exemption to dilution under section (3)(b) of the Lanham Act, which states that “[a]ll forms of news reporting and news commentary” are exempt from prosecution on account of dilution. [11] In this defense, OpenAI would argue that its use of the diluted trademarks constitutes “news commentary,” as ChatGPT provides analyses of news reporting by The New York Times. However, this argument depends on the outcome of the plaintiffs’ copyright claims, which would prove that OpenAI’s use of journalistic content does not constitute transformative analysis. This is because, in their copyright infringement case, the plaintiffs argue that OpenAI outputs referencing their newspapers are the result of memorization, where OpenAI’s LLM returns content memorized from a corpus containing content by the plaintiffs with some slight alterations, many of which are hallucinations or misrepresentations of their journalism. [12] Because news commentary requires transformative analysis and not simply a reproduction of published content, the news exemption would not be a valid argument if the plaintiffs win their copyright claims.
A second defense for Microsoft exists under the doctrine of fair use. Even if The Times loses their copyright case, OpenAI’s trademark dilution would not be protected under fair use, a doctrine that shields works from copyright and trademark claims on the basis of their commercial use, expressive nature, reliance on protected IP, and value. This is because the exemption on account of fair use applies to dilution “other than as a designation of source for the person’s own goods or services.” [13] However, the blurring and tarnishment that constitutes dilution occurs because of the improper designation of hallucinated material to the diluted trademarks, so this exemption would not be germane to the case. In Jack Daniel’s Properties v. VIP Products, the Supreme Court ruled on this very issue, where Jack Daniel’s, a whiskey company, sued VIP Products for trademark dilution because they had associated the product source with dog excrement in toy form, thus threatening likely harm to the reputation of the Jack Daniel’s trademark. In the case, the court sided with Jack Daniel’s, who argued that VIP Products’ satirical “use of ‘Bad Spaniels’ as a source identifier rendered its constitutional protections moot,” as “the Lanham Act’s ‘fair use’ defense … does not apply when the use designates a product’s source.” [14] Since The Times’ argument parallels Jack Daniel’s’ vis-à-vis the form of trademark dilution, OpenAI’s defense under the news exemption does not hold.
OpenAI has likely diluted the plaintiffs’ trademarks by blurring the distinctiveness of their trademarks through the attribution of the content of other papers and hallucinated materials to the diluted trademarks, and tarnishing their reputations by associating their newspapers with low-quality LLM outputs. Insofar as the output by ChatGPT has diluted the plaintiffs’ trademarks, and fair use is not a valid defense for OpenAI, the only remaining question is whether OpenAI can be held liable for tarnishment and blurring that ChatGPT caused. Based on precedent from Thomson Reuters v. Ross Intelligence, where Ross used Thomson Reuters’ catalogue of legal headnotes as part of their AI training corpus and Thomson Reuters was granted partial summary judgment on their copyright infringement claims, it is clear that AI companies, which control and garner profit from their LLMs, can be held legally accountable for infractions resulting from outputs of their LLMs, such as copyright infringement or trademark dilution. [15] Therefore, OpenAI can be held liable for trademark dilution, even if that dilution was performed by an AI model of the company.
The outcome of The Times v. Microsoft has enormous implications for both AI litigation and the future of journalism. This case will shape new legal frameworks for claims surrounding IP infringement from LLM outputs referencing the data of its training corpus; and with regard to its impact on journalism, the resolution of this case is inherently tied to public perception of the legitimacy of AI apropos of the news, and the necessity for traditional newspapers if LLMs can represent their journalism with the degree of accuracy required to defeat tarnishment claims. While The Times v. Microsoft is at the forefront of litigation regarding intellectual property violations by AI companies, it is just one of a series of trademark dilution cases where LLMs have allegedly blurred and tarnished marks through misattributions and hallucinations. In a similar case, Getty Images v. Stability AI, Stability AI generated images with Getty watermarks that actually lacked association with Getty Images, which Getty argued constituted trademark dilution. Here, Getty claimed that Stability AI’s “unauthorized use of the Getty Images Marks in connection with lower quality synthetic images” diluted the reputation of Getty’s brand. [16] More directly related to The Times case vis-à-vis journalistic credibility, in Dow Jones & Co. v. Perplexity AI, Dow Jones & Co., a media company owned by News Corp has alleged “false designation of origin and dilution” caused by Perplexity’s LLM. According to Dow Jones, Perplexity’s AI model generated “made-up text (hallucinations) in its outputs and attribut[ed] that text to Plaintiffs’ publications using Plaintiffs’ trademarks.” [17] Like The Times, Dow Jones. has achieved preliminary success against the defendant’s motion to dismiss. Due to the lack of case law apropos of dilution by AI-generated content, each of these cases will likely take several years to resolve, and their stakes — financially, and for the sake of journalism — demonstrate the imperative need for the application of existing trademark doctrines to the emerging field of AI.
Because The Times has demonstrated proof of likely damage to the diluted trademarks through documentation of erroneous misattributions of non-existent journalism to their newspaper, it is clear that public opinion regarding the credibility represented by the trademarks may reasonably be soured upon the association of hallucinated content to the newspapers. Furthermore, since this dilution is not protected under fair use or the news exemption of section 1125(c) of the Lanham Act, and because The Times has successfully defended itself against OpenAI’s motion to dismiss due to the lack of fame of the diluted trademarks, the court should grant injunctive relief to The Times to remedy the dilution caused by OpenAI.
Edited by Brandon Sahly and Ashley Zhou
[1] Seetharaman, Deepa, Stephen Nellis, and Deborah Mary Sophia. “Microsoft, OpenAI Reach New Deal to Allow OpenAI to Restructure.” Reuters, October 28, 2025. https://www.reuters.com/business/microsoft-openai-reach-new-deal-allow-openai-restructure-2025-10-28/.
[2] Hayes, Amanda. “What Is Trademark Dilution?” Nolo. Updated May 15, 2024. https://www.nolo.com/legal-encyclopedia/what-trademark-dilution.html.
[3] Act of July 5, 1946, 60 Stat. 427 (1946), https://www.govinfo.gov/content/pkg/COMPS-
1624/pdf/COMPS-1624.pdf
[4] The New York Times Company v. Microsoft Corporation et al., No. 1:23-cv-11195 (S.D.N.Y. Apr. 4, 2025), https://cases.justia.com/federal/district-courts/new-york/nysdce/1%
3A2023cv11195/612697/514/0.pdf?ts=1743868265
[5] Federal Trademark Dilution Act of 1995, Pub. L. No. 104-98, 109 Stat. 985 (1995), https://www.wipo.int/wipolex/en/legislation/details/14806
[6] Beerline, Jennifer Files. “Anti-Dilution Law, New and Improved: The Trademark Dilution Revision Act of 2006.” Berkeley Technology Law Journal 23, No. 1 (2008): 511-536. Page 512.
[7] “Lanham Act — Trademark Infringement and Dilution — First Amendment and Statutory Defenses — Jack Daniel’s Properties, Inc. v. VIP Products LLC.” Harvard Law Review, November 2023. https://harvardlawreview.org/wp-content/uploads/2023/11/
137-Harv.-L.-Rev.-440.pdf.
[8] Jaźwińska, Klaudia, and Aisvarya Chandrasekar. “How ChatGPT Search (Mis)Represents Publisher Content.” Columbia Journalism Review, November 27, 2024. https://www.cjr.org/tow_center/how-chatgpt-misrepresents-publisher-content.php.
[9] The New York Times Company. Complaint. December 27, 2023. pp. 52–54. https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf
[10] Borella, Michael. “The New York Times Case Against OpenAI Is Different. Here’s Why.” Patent Docs, February 19, 2024. https://www.patentdocs.org/2024/02/the-new-york-times-case-
against-openai-is-different-heres-why.html.
[11] U.S. Code. “15 U.S.C. § 1125(3)(b).” https://www.dmlp.org/sites/citmedialaw.org/
files/15USC1125.pdf
[12] Cooper, A Feder, and James Grimmelmann. “The Files Are in the Computer: On Copyright, Memorization, and Generative AI.” Chicago-Kent Law Review 100, no. 1 (2025): 188. https://doi.org/https://scholarship.kentlaw.iit.edu/cgi/viewcontent.cgi?article=4484&context=cklawreview.
[13] U.S. Code. “15 U.S.C. § 1125(3)(a).” https://www.dmlp.org/sites/citmedialaw.org/files/15USC1125.pdf
[14] Winshel, Alec. “Free Speech Is a Funny Thing: Jack Daniel’s Properties v. VIP Products Narrows First Amendment Protections for Trademark Usage.” Harvard Journal of Sports and Entertainment Law, December 28, 2023. https://journals.law.harvard.edu/jsel/2023/12/
free-speech-is-a-funny-thing-jack-daniels-properties-v-vip-products-narrows-first-amendment-protections-for-trademark-usage/.
[15] Centre GMBH and West Publishing Corp. v. Ross Intelligence, Inc., No. 1:20-cv-613-SB (D. Del Feb. 11, 2025), https://www.ded.uscourts.gov/sites/ded/files/opinions/20-613_5.pdf.
[16] Getty Images (US), Inc. v. Stability AI, Inc., No. 1:23-cv-00135-UNA (D. Del Feb. 2, 2023), p. 31, https://cases.justia.com/federal/district-courts/delaware/dedce/1:2023cv00135/
81407/1/0.pdf?ts=1682222407.
[17] Dow Jones & Company, Inc, and NYP Holdings, Inc. v. Perplexity AI, Inc., No. 1:24-cv-07984-KPF (S.D.N.Y. Aug. 21, 2025), pp. 4, 22, https://cases.justia.com/federal/district-
courts/new-york/nysdce/1:2024cv07984/630270/65/0.pdf?ts=1755867382