Cohere—Major news publishers sued Cohere alleging large-scale unauthorized use of copyrighted articles to train and serve language models
On February 13, 2025 a coalition of major US news publishers including Condé Nast, The Atlantic, Forbes, The Guardian, Los Angeles Times, McClatchy, Politico, The Republican Company, Toronto Star and Vox Media filed a copyright infringement lawsuit against Cohere in the Southern District of New York. The complaint alleged Cohere ingested and reproduced full articles from at least 4,000 publications and that the company's hallucinations falsely attributed fabricated content to named publishers. The suit was among the largest publisher-led actions against an enterprise AI company.
Scoring Impact
| Topic | Direction | Relevance | Contribution |
|---|---|---|---|
| Creator Compensation | -against | primary | -1.00 |
| Intellectual Property Ethics | -against | primary | -1.00 |
| Press Freedom | -against | secondary | -0.50 |
| Overall incident score = | -0.828 | ||
Score = avg(topic contributions) × significance (high ×1.5) × confidence (0.66)
Evidence (2 signals)
Judge denied Cohere's partial motion to dismiss publishers' copyright suit, allowing core claims to proceed
Cohere moved (May 22, 2025) to dismiss most counts of the publishers' copyright suit, arguing it is an enterprise B2B AI company rather than a consumer chatbot maker, that plaintiffs misused a developer demo tool, and that summarizing unprotected facts is not infringement. Judge Colleen McMahon (S.D.N.Y.) denied Cohere's partial motion to dismiss on November 13, 2025, allowing the core infringement claims to proceed.
Major news publishers filed copyright lawsuit against Cohere alleging unauthorized article ingestion
Wired and Reuters reported the February 13, 2025 filing by Condé Nast, The Atlantic, Forbes, The Guardian, Los Angeles Times, McClatchy, Politico, The Republican Company, Toronto Star and Vox Media against Cohere in the Southern District of New York, alleging large-scale unauthorized use of copyrighted articles to train and serve enterprise language models, plus hallucinated content falsely attributed to named publishers.