IP Law Daily, COPYRIGHT—S.D.N.Y.: Investigative news outlets’ DMCA claims against OpenAI dismissed without prejudice, (Nov 8, 2024)

By Saurabh Kashyap, B.A., LL.B., LL.M.

News outlets failed to demonstrate a concrete injury-in-fact and, thus, lacked Article III standing to pursue claims for damages or injunctive relief.

A federal district court in New York City has dismissed Digital Millennium Copyright Act claims brought by two online investigative news outlets against OpenAI, Inc. and its affiliated entities. The court found that the plaintiffs failed to demonstrate a concrete injury-in-fact resulting from the alleged removal of copyright management information in connection with training datasets for ChatGPT and, thus, lacked the required Article III standing to pursue claims for damages or injunctive relief. Additionally, the court denied the plaintiffs' motion for leave to replead without prejudice, allowing them to renew their motion by filing a notice of motion with a proposed amended complaint and an explanation demonstrating the viability of the proposed amendment (Raw Story Media, Inc. v. OpenAI, Inc., No. 1:24-cv-01514-CM (S.D.N.Y. Nov. 7, 2024)).

Background. The plaintiffs, Raw Story Media, Inc., and AlterNet Medis, Inc., are news organizations known for producing extensive investigative journalism, opinion articles, and breaking news, contributing over 400,000 articles collectively. The defendants, OpenAI, Inc., and affiliated entities (collectively, “OpenAI”), specialize in artificial intelligence and are recognized for the development of ChatGPT, an advanced language model trained on large-scale datasets to generate responses to user prompts.

The plaintiffs filed this lawsuit alleging that thousands of its copyrighted works were included in OpenAI’s training datasets without proper attribution or authorization, thus violating Section 1202(b)(i) of the Digital Millennium Copyright Act (DMCA). According to the plaintiffs, OpenAI's training process involved stripping these works of critical copyright management information (CMI), such as author names and titles, preventing the model from accurately citing the sources. OpenAI responded with a motion to dismiss, citing a lack of standing and, in the alternative, for failure to state a claim.

Standing and damages. In their argument for damages, the plaintiffs contended that the unauthorized removal of CMI constituted a tangible injury akin to interference with property under 17 U.S.C. § 1202(b)(i). They claimed that this injury paralleled common-law property rights, warranting recognition as a concrete harm. However, the court, referencing TransUnion LLC v. Ramirez, 594 U.S. 413 (2021), determined that the plaintiffs had not demonstrated any actual adverse effects from the removal of CMI, such as dissemination or economic loss, required to meet the standard for standing.

The court further noted that while Section 1202(b)(i) aims to prevent unauthorized tampering with CMI, it does not extend to protecting the exclusive rights tied to reproduction or derivative works as specified under 17 U.S.C. § 106. The decision cited Mango v. BuzzFeed, Inc., 970 F.3d 167 (2d Cir. 2020), emphasizing that the statute’s double-scienter requirement—that the defendant must both knowingly remove CMI and understand that such removal would facilitate infringement—was not fulfilled by the plaintiffs’ general assertions.

Injunctive relief. The plaintiffs also sought injunctive relief, requesting an order for OpenAI to remove any unauthorized uses of its works from its training datasets. They argued that there was a significant risk of future output from ChatGPT reproducing its articles without the proper CMI. The court, however, held that while removal of CMI could theoretically support injunctive relief, the plaintiffs did not demonstrate that the risk of such future harm was substantial or imminent, as required by Clapper v. Amnesty Int’l USA, 568 U.S. 398 (2013).

The court pointed out that the plaintiffs’ arguments were speculative, lacking concrete examples of actual incidents where ChatGPT reproduced its content verbatim. Highlighting the scale of data included in ChatGPT’s training corpus—comprising content from numerous sources—the court concluded that the likelihood of reproducing specific articles was minimal. This analysis referenced Pennell v. City of San Jose, 485 U.S. 1 (1988), underscoring the necessity for a “realistic danger” of direct harm for injunctive relief.

Conclusion. Ultimately, the court dismissed the plaintiffs’ copyright claims and denied their request for leave to amend without prejudice, allowing them to renew their complaint by filing a notice of motion with a proposed amended complaint and an explanation of why the amendment would not be futile.

The Case is No. 1:24-cv-01514-CM.

Judge: McMahon, C.

Attorneys: Kyle D. Wallenberg (Niro McAndrews, LLP) for Raw Story Media, Inc. and AlterNet Media, Inc. Allison Levine Stillman (Latham & Watkins LLP) for OpenAI Inc.

Companies: Raw Story Media, Inc.; AlterNet Media, Inc.; OpenAI Inc.; OpenAI GP, LLC; OpenAI, LLC; OpenAI Opco LLC

MainStory: TopStory AINews Copyright TechnologyInternet NewYorkNews GCNNews