NEW YORK: Encyclopaedia Britannica and Merriam-Webster have sued OpenAI in Manhattan federal court, accusing the artificial intelligence company of using nearly 100,000 online encyclopedia and dictionary entries without permission to train ChatGPT. The complaint, filed on March 13 in the Southern District of New York, says OpenAI copied copyrighted reference material at scale and used it in systems that generate answers for paying and nonpaying users. The plaintiffs are seeking monetary damages, a permanent injunction and a jury trial.
The lawsuit names multiple OpenAI entities, including OpenAI Inc., OpenAI LP, OpenAI LLC and OpenAI Group PBC. Britannica and Merriam-Webster allege copyright infringement and trademark violations, saying ChatGPT outputs reproduced or closely tracked their material and at times attributed inaccurate or incomplete content to their brands. The complaint says those uses could mislead readers into believing Britannica or Merriam-Webster approved, sponsored or supplied material that the companies say was generated without authorization.
In the filing, the plaintiffs cited examples they say show near-verbatim copying. One example involves Britannica’s article on education, another involves its article on tourism, and another concerns Merriam-Webster’s definition of the word “plagiarize.” The complaint also points to a ChatGPT response about the Hamilton-Burr duel that it says reproduced Britannica’s selection and ordering of quoted material. Britannica says those examples are demonstrative and that the full scope of any copying is within OpenAI’s control and records.
Complaint Cites Output Examples
Britannica says the alleged conduct has commercial consequences because AI-generated summaries can substitute for visits to Britannica and Merriam-Webster websites that depend on readers, subscribers and institutional users. The company describes itself in the complaint as a digital education and information platform built on continuously updated reference content for students, teachers and general readers. Merriam-Webster, also named as a plaintiff, is cited in the filing as a longstanding dictionary publisher whose copyrighted definitions and other entries are part of the works at issue.
OpenAI said Monday that its models are trained on publicly available data and grounded in fair use. The case adds to a growing docket of copyright disputes over generative AI training and output, and the complaint notes that related claims against OpenAI are already being handled in the same federal district through a multidistrict litigation. Britannica also filed a separate lawsuit against Perplexity last year, alleging similar misuse of its copyrighted reference material and trademarks in AI-generated answers.
Relief Sought In Manhattan Court
The complaint asks the court to award statutory damages, actual damages, restitution of profits, costs and attorneys’ fees, and to permanently bar the conduct described in the filing. It also demands a jury trial on all triable issues. Britannica and Merriam-Webster say OpenAI’s systems do more than summarize general knowledge, alleging that the company copied protected works for model training and for retrieval processes used to generate responses, then returned text that in some instances mirrored or closely followed the original material.
Filed under case number 1:26-cv-02097, the lawsuit places two long-established U.S. reference publishers in direct litigation with OpenAI over how copyrighted factual content is used inside generative AI products. For Britannica and Merriam-Webster, the case centers on whether copyrighted reference works and brand names were used without consent inside ChatGPT and related systems, and whether those outputs displaced visits to their own platforms. The case will proceed in the U.S. District Court for the Southern District of New York. – By Content Syndication Services.