A group of prominent publishers, including Advance Local Media LLC, Conde Nast, The Atlantic Monthly Group LLC, and others, have sued Cohere Inc., an AI company specializing in enterprise solutions, alleging it used their copyrighted works and trademarks without authorization to develop its large language models (LLMs) (Advance Local Media LLC et al. v. Cohere Inc., No. 1:25-cv-01305-CM, S.D.N.Y.).
Key Issues in the Complaint
The plaintiffs claim that Cohere violated their copyrights by using their works in training its LLMs and by generating outputs that allegedly include “verbatim copies,” “substantial excerpts,” or “substitutive summaries” of their works. They also allege secondary copyright liability, asserting that Cohere is responsible for any infringing outputs generated by its users. Additionally, the plaintiffs bring trademark infringement claims, arguing that Cohere’s AI tools “hallucinate” outputs that falsely attribute content to their publications, potentially causing consumer confusion.
Defendant’s Motion to Partially Dismiss
Cohere has filed a motion to dismiss several claims in the case, arguing that the plaintiffs’ allegations are insufficient and fail to meet legal standards. The defendant’s key arguments are as follows:
Secondary Copyright Liability: Cohere contends that the plaintiffs have not adequately alleged any instance of direct infringement by its users, which is a prerequisite for secondary liability. The company emphasizes that the examples of allegedly infringing outputs were generated by the plaintiffs themselves using Cohere’s demo tools, not by real-world users. Cohere also highlights that its terms of service prohibit uses that infringe intellectual property rights.
Direct Copyright Infringement (Substitutive Summaries): Cohere argues that summarizing or paraphrasing factual information does not constitute copyright infringement. The company asserts that the outputs cited by the plaintiffs do not copy protected expression but instead convey unprotected facts in original language.
Trademark Claims: Cohere challenges the plaintiffs’ trademark allegations, stating that the purported uses of their marks are not “in commerce” and do not plausibly cause consumer confusion. The defendant also argues that any use of the marks constitutes nominative fair use, as the marks are used solely to identify the plaintiffs’ publications in response to user prompts.
Further, Cohere maintains that the plaintiffs have failed to present plausible allegations of real-world infringement or consumer confusion. The company asserts that the plaintiffs’ claims are based on hypothetical scenarios and misuse of its demo tools, rather than actual conduct by its enterprise customers. Cohere adds that the publishers’ theories of liability would dangerously expand copyright and trademark law into areas that could stifle innovation and the dissemination of knowledge.
Conclusion
Cohere seeks to narrow the scope of the litigation by dismissing the claims for secondary copyright liability, trademark infringement, and direct copyright infringement based on “substitutive summaries.” The company emphasizes that the case should focus on the central issue of whether its training of LLMs constitutes fair use under copyright law. The court’s decision on this motion will likely shape the trajectory of this high-profile case involving generative AI and intellectual property rights.