IP Law Daily, COPYRIGHT—S.D.N.Y.: DMCA claim over ChatGPT’s ingestion of materials states an injury in fact, (Feb 21, 2025)

By Matthew Hersh, J.D.

The decision is the latest in a brewing debate over constitutional standing for copyright and DMCA claims.

An online newspaper could pursue its claims under the Digital Millennium Copyright Act against the owners of ChatGPT because it adequately stated an actual injury in fact, the federal court for Manhattan has held. The court’s opinion, which green-lit the lawsuit even though it was based only on ChatGPT’s ingestion of materials for training, was the latest entry—and will likely not be the last—over the application of Article III standing claims to copyright claims that do not allege actual distribution of the allegedly infringing work (The Intercept Media, Inc. v. OpenAI, Inc., No. 1:24-cv-01515-JSR (S.D.N.Y. Feb. 20, 2025)).

The lawsuit is one of many to take on major artificial intelligence platforms over their unauthorized ingestion of copyrighted material. The Intercept is a progressive-leaning publication with a significant online presence. The media company filed a lawsuit in February 2024 against OpenAI and Microsoft based on ChatGPT’s alleged use of its online articles as part of its training set. The complaint alleges that OpenAI and Microsoft infringed its copyrights by copying the materials when ingesting the works as part of its training process. The complaint also alleges that ChatGPT violated the DMCA by stripping from its training dataset key information—described under the statute as content management information, or CMI—identifying the authorship of the works.

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

Standing. The court, in an opinion by longstanding Manhattan judge Jed Rakoff, found that The Intercept had standing to pursue its DMCA claim. The dispute between the parties was whether the news publication had sufficiently alleged the “concrete harm” necessary to have standing. In Transunion v. Ramirez, 94 U.S. 413 (2021), the Supreme Court’s latest brush with the issue, the Court explained that the central inquiry came down to “whether the alleged injury to the plaintiff has a close relationship to a harm traditionally recognized as providing a basis for a lawsuit in American courts.” OpenAI and Microsoft argued that the DMCA claim in this case did not have such a relationship. The district court disagreed.

The DMCA claim met the test, the court concluded, because it sought to vindicate a similar type of property interest as a traditional copyright claim. To be sure, the court acknowledged, copyright claims protect against infringements of the exclusive rights over works granted by the Copyright Act “while DMCA claims only seek to protect those works’ CMI.” But the question was not the specific claim anticipated by the statute, the court emphasized, but “whether there is a close historical or common-law analogue for The Intercept’s asserted injury.” And here, the court noted, the Copyright Act formed the analogue itself. Like the Copyright Act, the court reasoned, the DMCA was intended to protect a right historically protected under the under law, namely property rights. And just like the Copyright Act, the court noted, the property rights protected under the DMCA are designed to incentivize creative production. “To state the conclusion clearly,” the court emphasized, “even though the specific right created by the DMCA may be comparatively new, the injury experienced by The Intercept because of the violation of that right sounds in the same kind of harm long recognized in copyright suits.

Nor did it matter, the court noted, that the news publication had failed to identify a public dissemination of its copyrighted works. The ChatGPT parties again relied on Transunion for this point, but the case was again distinguishable. There, a class of credit holders sued a credit reporting agency for including inaccurate information in their credit report. The Court found that only those class members whose allegedly inaccurate information had actually been disseminated to creditors could allege an injury in fact. ChatGPT pressed the analogy here, but the court rejected it. “TransUnion,” the court noted, “dealt with an analogy to the common-law tort of defamation.” But “[a] copyright injury—or, more generally, an injury to a property right,” the court reasoned, “does not require publication to a third party.” Thus, the DMCA claim could go forward.

Failure to state a claim. But while the ChatGPT parties would fail to earn dismissal of the DMCA claims case on standing grounds, they had more success on the merits. The claims at issue were brought under two different provisions of the DMCA. One, arising under Section 1202(b)(1) of the Act, claimed that ChatGPT carved out CMI when ingesting The Intercept’s news articles into its training model. The other, arising under Section 1202(b)(3) of the Act, alleged that ChatGPT knowingly distributed materials to Microsoft—and vice versa—without that CMI. The first claim would go forward, the Court found, but the second would not.

The claim under Section 1202(b)(1) could go forward against OpenAI, the court found, because The Intercept plausibly alleged all of the necessary elements. The Intercept plausibly alleged that OpenAI intentionally removed CMI from its articles because it identified the specific training sets—training sets that originally contained CMI before OpenAI extracted it—that it claimed OpenAI used to train ChatGPT. The tougher question, the court noted, was whether OpenAI excluded CMI from its training with knowledge that doing so would “induce, enable, facilitate, or conceal an infringement.” But the news publication met this threshold as well, the court found, because it specifically pleaded that some ChatGPT users might have been deterred from using outputs that contained the publication’s CMI due to copyright infringement concerns. Moreover, the court noted, The Intercept also alleged that OpenAI indemnified ChatGPT users from copyright claims—clear evidence, the court reasoned, that “the company knew some users would likely infringe copyright.”

But while the Section 1202(b)(1) claim against OpenAI would go forward, the court found, the claim against Microsoft would not. Critically, the court noted, none of the specific items of evidence related to Microsoft's involvement with ChatGPT had any relationship to alleged CMI removal. Rather, the court noted, “all the specific factual matter in the complaint related to CMI removal connects only to OpenAI.” Microsoft would be relieved of this claim.

Finally, the claim under Section 1202(b)(3) would fail against both parties. The problem, the court noted, was that the complaint included no factual support for its allegation that Microsoft and OpenAI actually distributed its articles. The only distribution that the complaint alleged, the court noted, was an allegation that Microsoft and OpenAI had shared copies of CMI-stripped works with one another. But the factual support for these allegations, the court noted, derived exclusively from “general observations about the business relationship between OpenAI and Microsoft.” That was not sufficiently specific, the court emphasized, to allow the claim to go forward.

Comment. Judge Rakoff was not the first to grapple with the application of Transunion to DMCA claims—although he charted a different course than a colleague on the same court. In Raw Story Media, Inc. v. OpenAI, Inc., a DMCA case also pending in the Southern District, Judge Colleen McMahon concluded that DMCA plaintiffs did not have standing if they did not also allege that the CMI-stripped works were actually distributed to the public. Judge McMahon’s opinion immediately drew widespread attention in the copyright community for its possible implications even for analogous claims under the Copyright Act.

The Case is No. 1:24-cv-01515-JSR.

Judge: Rakoff, J.

Attorneys: Jon Loevy (Loevy & Loevy) for The Intercept Media, Inc. Allison Levine Stillman (Latham & Watkins LLP) for OpenAI, Inc.

Companies: The Intercept Media, Inc.; OpenAI, Inc.; OpenAI GP, LLC; OpenAI, LLC; OpenAI Opco LLC; OpenAI Global LLC; OAI Corporation, LLC; OpenAIHoldings, LLC; Microsoft Corporation

MainStory: TopStory AINews Copyright TechnologyInternet NewYorkNews GCNNews