Key Points

1. A large number of news publishers (Publishers) previously sued AI company Cohere for copyright and trademark infringement

   
   

2. Cohere just brought a partial motion to dismiss the lawsuit, and it lost—on November 13, 2025, McMahon, J of the United States District Court Southern District of New York denied Cohere’s partial motion to dismiss the lawsuit

   
   

3. This dispute is one of more than 50 lawsuits that are currently before the courts challenging the use of copyrighted works by AI companies to train their large language models—each case depends on the circumstances, and we will have to wait and see

Back in March, 2025, I wrote about how several news publishers (Publishers) sued AI company Cohere for copyright and trademark infringement.

As a refresher, the Publishers alleged that Cohere, without permission or compensation, used scraped copies of their articles, through training, real-time use, and in outputs, to power its AI service, which in turn competed with Publisher offerings and the emerging market for AI licensing. The Publishers accused Cohere of stealing their works to the point where actual verbatim copies were produced in outputs, and of blatantly manufacturing fake pieces and attributing them to the Publishers, which misled the public and tarnished their brands. What’s more, when RAG was turned off, the AI model, Command, hallucinated answers. Ultimately, the Publishers claimed that Cohere’s actions amounted to “massive, systematic copyright infringement and trademark infringement, and have caused significant injury to Publishers”.

Well now, a new development has emerged—Cohere brought a partial motion to dismiss the lawsuit, and it lost. That is, on November 13, 2025, McMahon, J of the United States District Court Southern District of New York denied Cohere’s partial motion to dismiss Counts II, III, and IV of the Publishers' complaint.

Why Did the Court Deny Cohere’s Motion?

In this motion, the judge found the following:

• Cohere's Motion to Dismiss the Publishers' Direct Copyright Infringement Claim was Denied. Cohere wanted to dismiss the Publishers' claim for direct copyright infringement only to the extent it alleged that Cohere was directly liable for generating "substitutive summaries" of the Publishers' work. In a nutshell, Cohere argued that though the Publishers could show they owned valid copyrights, Command’s summaries were not substantially similar to the Publishers’ works. The court disagreed with Cohere and concluded that the Publishers adequately alleged that Command's outputs were quantitatively and qualitatively similar—they argued that Command's output heavily paraphrased and copied phrases verbatim from the source article, and that these summaries went well beyond a limited recitation of facts. Also, the Publishers provided 75 examples of Cohere's alleged copyright infringement (50 allegedly included verbatim copies and a further 25 examples had a mix of verbatim copying and close paraphrasing)

  
  

• Cohere's Motion to Dismiss the Publishers' Secondary Copyright Infringement Claim was Denied. The Publishers claimed that Cohere was secondarily liable for unlawfully reproducing, displaying, distributing, and preparing derivatives of the Publishers' copyrighted works under each of three theories: contributory infringement by material contribution, contributory infringement by inducement, and vicarious infringement. Cohere agued that all three theories failed, but the court held that each of Cohere’s arguments were without merit. In particular, the Publishers adequately alleged underlying direct infringement; the Publishers adequately alleged Cohere’s knowledge of direct infringement; and the Publishers adequately alleged inducement

  
  

• Cohere's Motion to Dismiss the Publishers' Lanham Act (trademark) Claims was Denied. Cohere argued that the Publishers failed to allege use of their marks in commerce and a plausible likelihood of consumer confusion and that Cohere's use of the marks was lawful as nominative fair use. However, the court disagreed with Cohere because the Publishers adequately alleged Cohere’s use in commerce; the Publishers adequately alleged a likelihood of confusion; and the nominative fair use doctrine did not apply on the facts of this case—"All I can and will do is conclude that the complaint adequately alleges facts that could, if proved, cause a trier of fact to reject application of that doctrine”

Needless to say, the court was simply unconvinced by Cohere’s arguments and shot them all down. Since Cohere was unsuccessful, it will have to prepare for a trial.

What Does This Mean for the Case?

This is not good for Cohere. At this point in the case, it is striking that the Publishers put Cohere on notice that it was not allowed to do what it was doing—the Publishers had copyright notices and terms of service on their websites, and they also sent do-not-crawl instructions to Cohere’s bots using robots.txt protocols. In fact, the Publishers also claimed that Cohere kept unlawfully copying their works even after they sent a cease-and-desist letter.

From the perspective of the Publishers, this motion went well; if this is an indication of what is to come later down the line, they will be content.

What Does This Mean for the Other AI Infringement Cases?

The court noted that this dispute is one of more than 50 lawsuits that are currently before the courts challenging the use of copyrighted works by AI companies to train their large language models.

Some may view the decision as a foreshadowing of what could transpire in some of the other cases, but it is important to note that this was just one motion in one case; the decisions of those other cases will depend on the circumstances of those cases. We can only wait and see.