IP Law Daily, COPYRIGHT—S.D.N.Y.: Media company claims over ChatGPT are narrowed, but mostly go forward, (Apr 7, 2025)

By Matthew Hersh, J.D.

The companies adequately stated violations of copyright, trademark, and DMCA violations.

Three prominent news media companies adequately stated claims under the Copyright Act, the Lanham Act, and the Digital Millennium Copyright Act against the two companies behind ChatGPT, the federal court for Manhattan has held. The highly anticipated opinion, covering more than 15,000 words of the Federal Reporter, provided a treatise of sorts on the statutes at issue—although it is not likely to be the last word in the extensive and ongoing nationwide litigation over artificial intelligence technology and their implications for traditional copyright industries (New York Times Co. v. Microsoft Corp., No. 1:23-cv-11195-SHS-OTW (S.D.N.Y. Apr. 4, 2025)).

.The opinion comes in one of the many lawsuits pending nationwide against Microsoft and OpenAI, the companies behind the popular ChatGPT artificial intelligence platform. The action was originally made up of three different lawsuits against the ChatGPT operators—one brought by The New York Times, another brought by The Daily News, and the third brought by the Center for Investigative Reporting. The three lawsuits were eventually consolidated into a single action in front of Senior Judge Sidney Stein, a 30-year veteran of the federal bench. The consolidated cases present a wide range of claims under the Copyright Act, the Digital Millennium Copyright Act, the Lanham Act, and several other state and common law provisions.

Microsoft and OpenAI moved to dismiss the complaint, leading to this extensive opinion.

Statute of limitations. The court first rejected Microsoft and OpenAI’s arguments—brought only against the New York Times and Daily News—on the Copyright Act statute of limitations. Because the media companies filed their lawsuits in 2023 and 2024, respectively, the court noted, their claims would be time-barred if they “discovered, or with due diligence should have discovered” the alleged infringement in 2020 and 2021. Microsoft and OpenAI argued that the media companies knew or should have known, prior to that time period, that ChatGPT was being trained on their content. The court disagreed.

The court rejected the statute of limitations argument because the evidence Microsoft and OpenAI tendered—at least at this early stage of the lawsuit—was not sufficient to establish the defense. The companies came forward with a slate of news reports from that time period describing how their artificial intelligence models were training on massive sets of data—as one 2020 article put it, “billions of pages of publicly available documents on the internet.” But these articles were not enough, the court found, because they did not make it clear that the media companies’ particular claims should have been discovered. Moreover, the court noted, OpenAI did not became a household name until “the release of ChatGPT in November 2022.”

In rejecting the statute of limitations claim, the court also emphasized that it would not treat the media companies any differently than it would any other plaintiff. Microsoft and OpenAI argued that because the companies were sophisticated plaintiffs, they should have been following more closely than other rightsholders ChatGPT’s use of their works. But the Second Circuit had rejected this “sophisticated plaintiff” doctrine, the court noted, so it would be no help to Microsoft and ChatGPT here. “Discovery may reveal facts supporting OpenAI’s contention that The Times and the Daily News plaintiffs discovered the alleged infringement before December 27, 2020 and April 30, 2021 respectively,” the court noted. But at the motion to dismiss stage, however, the court concluded, the companies failed to meet their burden.

Contributory copyright infringement. The court also refused to dismiss the claim of contributory copyright infringement. Microsoft and OpenAI argued, this time taking on all three of the plaintiffs, that the media companies had failed to plausibly allege both direct infringement by ChatGPT’s users and that defendants knew of third-party infringement. The court rejected both of these arguments.

With respect to copyright infringement by third party users, the court found, this was relatively easily established. The media companies came forward with collectively more than 100 examples in which users were able to obtain copyrighted content with specific prompts to ChatGPT—in some cases by simply asking the service to give them copies of specific copyrighted articles. These examples “raise a reasonable expectation,” the court noted, “that discovery will reveal evidence of additional examples of third-party infringement such that dismissal at this stage would be improper.”

The media companies also adequately pleaded that the companies behind ChatGPT had sufficient knowledge of that third party infringement to sustain a contributory infringement claim. While the circuits were split on the level of scienter required to establish a contributory copyright claim, the court noted, the standard in the Second Circuit is whether the defendants objectively “know or have reason to know” of the direct infringement by third-party end users. Under that standard, the court noted, the media companies had plausibly alleged that Microsoft and OpenAI possessed “constructive, if not actual, knowledge of end-user infringement.” The complaints alleged that the companies knew they were using copyrighted works to train their models and that there “widely publicized” reports that end users were using ChatGPT to elicit copyrighted content, the court noted. Indeed, the Times even informed Microsoft and OpenAI that their tools infringed its copyrighted works, the court noted, “supporting the inference that defendants possessed actual knowledge of infringement by end users.” The claim would go forward.

Nor could Microsoft or OpenAI seek refuge in the “substantial noninfringing uses” doctrine set forth in Sony Corp. of Am. v. Universal City Studios, Inc. and its progeny Metro-Goldwyn-Mayer Studios Inc. v. Grokster, Ltd. In Sony, the Supreme Court found that the makers of a product with substantial noninfringing uses—in that case, a VCR—were not liable for contributory material infringement if the product also had potential legitimate uses. In Grokster, the Court clarified that the distributor of such a product—in that case, an internet file-sharing app—could be liable if it “induced” copyright infringement by its users. Microsoft and OpenAI argued that under these cases, inducement was essentially a floor—in that words, that the distributor of a product with substantial noninfringing users could never be liable absent a finding of inducement. But that misread the two cases, the court found. The Sony doctrine merely foreclosed imputing culpability “solely based on the characteristics or uses of a distributed product,” the court noted. It left open, however, “other rules of fault-based liability derived from the common law, including liability based on material contribution”—the claim at issue here. The claims would go forward even without inducement allegations.

DMCA claims—standing. The court then turned to the media companies’ claims under the Digital Millennium Copyright Act. The court first had to determine whether the media companies had standing to bring their claims. The court found that they did.

The court first found that the media companies adequately alleged a cognizable harm under Article II. Microsoft and OpenAI argued that under Transunion v. Ramirez, a 2021 decision tightening the standard for cognizable harm, the media companies could not assert a DMCA violation merely because ChatGPT stripping their copyright management information from news articles when uploading them into its training module. But the harm the media companies suffered by the stripping of this CMI, the court found, was a sufficiently close analogue to harm traditionally suffered under the Copyright Act—making it sufficient, even under Transunion, to state a cognizable harm under Article III. (The question of the applicability of Transunion to cases arising under the DMCA is one that is arising with increasing frequency in recent months. IP Law Daily last covered the issue here).

The court also found that the media companies adequately pleaded the causation required to establish Article III standing. Here, the court noted, the media companies alleged that ChatGPT’s removal of copyright management information (CMI) from their works “conceals and facilitates copyright infringement, which deprives plaintiffs of licensing and subscription revenue.” That harm, the court emphasized, was “fairly traceable to the removal of CMI: its removal allows defendants to provide plaintiffs’ works directly to end users through regurgitating outputs, while concealing that defendants infringed plaintiffs’ copyrights to generate those outputs.” The DMCA claims would survive the Article III challenge.

DMCA claims—merits. The DMCA claims by the Daily News and the Center for Investigative Reporting would also go forward on their merits, the court found—but the claim by the Times would not. The Daily News and CIR offered detailed allegations about how ChatGPT’s systems worked, the nature of the manner in which CMI was allegedly stripped, and the fact that the media companies were publicly known to remove such CMI. But the Times complaint, by contrast, the court noted, “does not include any specific detail on how CMI was allegedly removed during the training process, and its conclusory statement that defendants’ process of training their LLMs removes CMI ‘by design’ fails to nudge its claims across the line from conceivable to plausible.”

As to the DMCA’s second scienter requirement, the court found, that threshold was also met by the pleadings. The media companies were required to plead not only that ChatGPT removed CMI, but that its makers knew or had reason to know that distribution of copyrighted material despite the removal of CMI would “induce, enable, facilitate, or conceal an infringement.” Here, the court noted, the complaints alleged that OpenAI “has publicly acknowledged both that it uses copyrighted works to train its models and that its models ‘are capable of distributing unlicensed copies of copyrighted works.’” The complaints also alleged, the court noted, that OpenAI “was aware of the possibility of end-user infringement through the generating of regurgitating outputs, including because of OpenAI’s policy of agreeing to indemnify end users accused of infringement.” The DMCA claims based on removal of CMI would go forward.

But although the claims under Section 1202(b)(1) would go forward, the court noted, the claims under Section 1202(b)(3)—prohibiting the distribution of CMI-less works—would not. The problem here was that under caselaw interpreting this provision of the DMCA, only the distribution of substantially complete CMI-less works—not merely portions of them—could be actionable. Reviewing the regurgitations cited in the Times and Daily News complaints, the court noted, the regurgitation did not meet that standard because they were generated in response to multiple prompts, and “outputted in a different order than they appear in the original articles, falling far short of representing exact or substantial reproductions of the originals.” They could not be the basis of a Section 1202(b)(3) claim.

Common law misappropriation. While the Copyright Act claims and most of the DMCA claims would go forward, the court found, the common law misappropriation claim would not. That claim, the court reasoned, rested on alleged actions that were similar to those alleged in the copyright claims—thus preempting those claims under the Copyright Act. The media companies argued that their misappropriation claims asserted an additional element was not part of the copyright claims, namely that ChatGPT was undermining their right to distribute breaking news to their readers. But this so-called “hot news” exception to the preemption doctrine did not apply here, the court found, because none of the excerpts attached to the media companies’ complaints satisfied “the time-sensitivity requirement of the hot news exception.” Indeed, the court noted, the media companies “nowhere allege that the ChatGPT outputs... regurgitated each article precisely at the point where the profit was to be reaped.” The exception therefore would not apply, the court found—and the claim would be preempted.

Trademark dilution. The court found that the media companies’ federal and state trademark claims would go forward. In these claims, brought only by the Daily News and several of its constituent papers, the companies alleged they were owners of “distinctive and famous marks” that are widely recognized by the general consuming public. Moreover, the companies alleged, ChatGPT had used those marks, without authorization, on “lower-quality and inaccurate writing,” thereby diluting the quality of the marks by tarnishment. The claims passed muster, the court found.

Most importantly, the court noted, the Daily News and its constituent media companies adequately pleaded that their marks were sufficiently famous under the Lanham Act. The companies alleged, among other things, that their publications had been in circulation for more than 100 years and circulated throughout all 50 states, their news stories were “featured by major national news outlets and have received significant attention and praise from news outlets such as CNN, MSNBC and Fox News,” their publications have “achieved national and international fame for their reporting of highly significant events in both U.S. and world history,” and their publications have “received widespread recognition for their achievements including numerous Pulitzer Prizes.” That was easily enough, the court found, to meet the standard required under the Lanham Act.

The claim under New York state law also would go forward, the court found. Microsoft and OpenAI rested principally on their argument that N.Y. Gen. Bus. Law § 360-l violated the Dormant Commerce Clause of the Constitution because it discriminated against out-of-state commerce. But the argument would fail, the court found. The New York dilution statute may “have a considerable influence on commerce outside New York’s borders,” the court noted, but the ChatGPT makers “have not contended much less shown that section 360-l discriminates against out-of-state commerce.” The New York claim, like the claim under its federal counterpart, would go forward.

Abridgements. Finally, the court noted, the Center for Investigative Reporting’s copyright infringement claims over ChatGPT’s outputs of abridgements of its articles—as opposed to full reproductions or excerpts—would fail. The alleged abridgments, the court noted. were “detailed summaries, usually in bullet point form, of the facts contained in CIR’s articles.” Those summaries, which “differ in style, tone, length, and sentence structure from CIR’s articles,” the court noted, were not substantially similar to CIR’s copyrighted works. Indeed, the court noted, they “present the facts in a different arrangement—bullet point lists or short summary paragraphs—with a different sentence structure and different phrasing.” That could not be the basis, the court concluded, for a copyright infringement claim.

The Case is No. 1:23-cv-11195-SHS-OTW.

Judge: Stein, S.

Attorneys: Adnan Muttalib (Susman Godfrey L.L.P.) for The New York Times Co. Jeffrey S. Jacobson (Faegre Drinker Biddle & Reath LLP) for Microsoft Corp.

Companies: The New York Times Co.; Microsoft Corp.

MainStory: TopStory Copyright TechnologyInternet Trademark NewYorkNews GCNNews AINews