A U.S. federal court has officially allowed a major copyright infringement lawsuit against Cohere Inc., the Canadian artificial intelligence startup valued at over $5 billion, to proceed to trial. The lawsuit, spearheaded by fourteen leading publishers—including Condé Nast, The Atlantic, Forbes, and The Guardian—centers on allegations that Cohere unlawfully used over 4,000 copyrighted works to train its AI models, resulting in widespread reproduction and distribution of proprietary publisher content across its commercial platforms.
The Lawsuit and Its Plaintiffs
Filed in February 2025, the lawsuit brings together 14 leading publishers, including Condé Nast, Forbes Media, The Atlantic, The Guardian, Toronto Star, McClatchy, Los Angeles Times, Vox, Politico, and others. The group alleges that Cohere engaged in widespread copyright and trademark infringement by systematically scraping and reproducing thousands of their articles to train its AI models and generate competing “substitutive summaries” without authorization or compensation. The publishers are seeking damages that could amount to $150,000 for every article unlawfully reproduced, distributed, or displayed.
Court’s Ruling and Reasoning
On November 13, 2025, Judge Colleen McMahon of the Southern District of New York denied Cohere’s motion to dismiss the case, rejecting arguments that its AI model simply repurposed “abstracted facts.” Instead, the court found credible evidence of direct copying, with Cohere’s Command AI model allegedly copying and pasting entire paragraphs verbatim from publisher content. The judge also recognized sufficient grounds on several fronts:
- Direct copyright infringement from both training and output activities,
- Secondary liability for alleged inducement of infringement,
Trademark infringement resulting from AI-generated outputs falsely attributed to publishers, - Misuse of retrieval-augmented generation (RAG) techniques to extract and reproduce articles in real-time,
- Continued unauthorized use despite “do-not-crawl” instructions, website copyright notices, and cease-and-desist letters from publishers.
The decision called Cohere’s arguments “without merit” and described the allegations of systemic infringement as warranting full legal review. Danielle Coffey, President and CEO of the News/Media Alliance, said, “We’re grateful that Judge McMahon soundly rejected all aspects of Cohere’s motion to dismiss. As the complaint alleges, Cohere has engaged in massive, systemic infringement, and affected publishers deserve their day in court.”
Industry and Legal Impact
The lawsuit is already being described as a “first step towards justice for publishers,” according to news media executives, as it addresses a central challenge posed by generative AI and content platforms, namely, the unlicensed reproduction of journalism and magazine articles, which threatens revenue streams tied to advertising and subscriptions. By allowing the case to proceed, the court has sent a clear message: copyright law applies to AI-generated outputs, and swift adaptation of legal frameworks is necessary as technology continues to evolve.
Publishers argue that platforms like Cohere diminish user incentives to visit original news sites, reducing traffic, subscriptions, and ad revenue. They also highlight reputational risks from fabricated AI-generated articles erroneously attributed to their brands. The suit mirrors other high-profile copyright disputes, notably against OpenAI, which saw its attempt to transfer jurisdiction to the United States rejected, reinforcing the principle that local copyright protections remain enforceable.
Response from Cohere
For its part, Cohere has consistently denied allegations, characterizing the lawsuit as “misguided and frivolous.” The company claims to train its AI models responsibly and maintain controls to mitigate IP risks, while expressing willingness to engage in direct dialogue with publishers to resolve their concerns. However, these assertions have not prevented the legal battle from escalating, as the judge’s decision highlights the legal and financial risks that AI startups face if copyright protections are overlooked.
Next Steps and Outlook
The case is now headed towards fact discovery and a potential jury trial, with publishers having until December 1, 2025, to file amended complaints and further outline their damages. Expert discovery is scheduled to run until July 2026, positioning the lawsuit as a focal point for future regulation of generative AI in the publishing industry.
