Arlington, VA – Earlier today, a collection of news and magazine publishers suing AI company Cohere for copyright and trademark infringement filed a brief opposing the company’s motion to dismiss parts of the case.
Danielle Coffey, President and CEO of the News/Media Alliance, the leading industry trade association that represents each plaintiff, said, “Cohere engages in widespread unauthorized use of publisher content in developing and running its generative AI systems. Our complaint identifies thousands of examples of outputs from the company’s AI product that demonstrate massive, systematic copyright infringement and trademark infringement. Cohere’s motion to dismiss seeks to dodge responsibility for this behavior, claiming that users don’t actually use AI tools to access news, even though its own marketing touts its ability to provide access to the latest articles. Today’s filing shows just how misguided Cohere’s argument is.”
The filing by the plaintiffs, a group including Advance Local Media, Condé Nast, The Atlantic, Forbes Media, The Guardian, Business Insider, LA Times, McClatchy Media Company, Newsday, Plain Dealer Publishing Company, POLITICO, The Republican Company, Toronto Star Newspapers, and Vox Media, strongly rejects the arguments laid forth in Cohere’s motion to dismiss, including the following:
On Cohere’s argument that it markets to enterprises, not individuals, and therefore its users wouldn’t use its product to obtain news articles:
“The Complaint and its exhibits contain extensive and detailed factual allegations regarding direct infringement. Rather than conduct its own journalism, Cohere supplies users with articles taken without consent from, or compensation to, the creators of the content. Cohere essentially acts as an unlicensed news syndicator. … Even when asked generally about a topic, Cohere provides Publishers’ articles.”
“Cohere’s users can be expected to use Command to obtain news and magazine articles, as Cohere specifically marketed that as a relevant use case.”
“News delivery is a key selling point for Cohere. In July 2023, Cohere’s first post announcing Chat showed the model retrieving an article from Wired, a Condé Nast publication. That September, Cohere again prodded users to access news articles through its models, suggesting that developers can use Command ‘to access the latest news about trends and competitors in their space.’ … Publishers’ examples depict Cohere’s service in action, with natural language prompts under the very “use case” that Cohere repeatedly encourages and expects.”
On Cohere’s effort to excuse copying for use in directly competitive, fully substitutional summaries:
“Cohere misunderstands the relevance of these summaries to evaluating Cohere’s repeated infringements and misrepresents well-settled case law on when summaries infringe. While Cohere is perfectly free to investigate and report the ‘facts,’ just as Publishers do when covering the news, Cohere enriching itself by copying journalistic expression is infringement.”
The brief filed today carefully walks through established case law, starting with Justice Story’s holding that non-fictional abridgements infringe when they comprise “the facile use of the scissors; or extracts of the essential parts, constituting the chief value of the original work.”
“Publishers’ Complaint details extensive overlap between the copyrighted works and the infringing substitutive summaries in organization, factual selection, emphasis, language, writing style, and overall feel.”
On Cohere’s claim that its use of publisher names and trademarks, even when providing incorrect information attributed to those publishers to its users, should be ignored:
“Publishers’ businesses and hard-won reputations hinge on decades of trustworthy journalism,” but “Cohere distributes fake articles bearing Publishers’ marks to third parties as part of its commercial services.”
“Cohere’s display of their trademarks on its AI-generated articles will likely divert traffic, sales, and subscriptions from Publishers. ‘Attribution makes users more likely to trust the response they got from Cohere, rather than causing the user to navigate to the original source.’ Thus, when Cohere delivers articles ‘falsely attributed to Publishers,’ consumers will tend to rely on Cohere to access Publisher’s articles, ‘rather than registering with, or subscribing to Publishers, or obtaining content from another licensed source.’”
The complaint provided a non-exhaustive list of thousands of articles that Cohere has infringed, through training, real-time use of content, and infringing outputs, listed below along with additional information about the case.
- Download the complaint, Ex. A (Works in Suit), Exh. B (Copyright infringements); Ex. C (Trademarks), Ex. D (Trademark Infringement).
- Read an op-ed on the case by Danielle Coffey, President and CEO of the News/Media Alliance
- Read the Wall Street Journal’s reporting on the case
