Last reviewed
Jun 3, 2026
Sources
9 citations
Review status
Source-backed
Revision
v1 · 1,491 words
Add missing citations, update stale details, or suggest a clearer explanation.
Britannica and Merriam-Webster v. OpenAI is a copyright and trademark lawsuit filed on March 13, 2026 in the United States District Court for the Southern District of New York. Encyclopaedia Britannica, the 250-year-old reference publisher, and its dictionary subsidiary Merriam-Webster sued OpenAI over the way ChatGPT was built and the way it answers questions. The complaint alleges that OpenAI copied close to 100,000 of Britannica's online articles to train and ground its large language models, that ChatGPT reproduces the publishers' text verbatim or near-verbatim, that it attaches the companies' famous brand names to fabricated answers, and that it "starves" the publishers of the web traffic and revenue that fund their work. The case is one of the clearest tests yet of whether AI "answer engines" can lawfully substitute for the reference sources they were trained on. [1][2][3]
The plaintiffs are Encyclopaedia Britannica, Inc., headquartered in Chicago, and Merriam-Webster, Inc., based in Springfield, Massachusetts. Britannica owns Merriam-Webster, and the two are referred to collectively in the filing as "Plaintiffs." Britannica describes itself as "a household name synonymous with trusted, fact-checked, meticulously researched content," and notes that Merriam-Webster has been "America's leading provider of language information for more than 180 years." Its Collegiate Dictionary is among the best-selling books in American history. [4]
The defendants are nine related OpenAI entities, from the original nonprofit OpenAI, Inc. through the for-profit OpenAI Group PBC and the operating companies that sell ChatGPT, ChatGPT Enterprise, and the OpenAI API. The case is captioned Encyclopaedia Britannica, Inc. v. OpenAI, Inc., No. 1:26-cv-02097, and was assigned to the Southern District of New York, the same court hearing New York Times v. OpenAI. The plaintiffs are represented by Susman Godfrey, the litigation firm that also represents The New York Times against OpenAI. They demanded a jury trial and signed the complaint on March 13, 2026. [4][5]
The complaint frames the dispute as one about substitution rather than search. A traditional search engine, the plaintiffs argue, is an intermediary: it takes a query and returns links that send users to the publishers who actually hold the information. Those clicks let publishers sell subscriptions and advertising. ChatGPT, by contrast, "absorbs" the content and hands the user a finished answer, so the reader never visits Britannica or Merriam-Webster at all. The filing puts it bluntly: ChatGPT "starves web publishers like Plaintiffs of revenue by generating responses to users' queries that substitute, and directly compete with, the content from publishers like Plaintiffs." [4]
Most of the complaint is about copyright. The plaintiffs say OpenAI infringed their rights at three stages. First, by copying their content "at massive scale" to train the LLMs behind ChatGPT. Second, by retrieving and copying that content again through retrieval-augmented generation, the technique that lets a chatbot pull in fresh material to supplement what the model memorized. Third, by generating outputs that contain "full or partial verbatim reproductions" of the articles, or that paraphrase, summarize, or mimic "the selection and curation of content and lists" in Britannica's work. [4]
The complaint includes specific examples that it says show copying. Asked "How does Merriam-Webster define plagiarize," ChatGPT returned the exact definition from the Merriam-Webster Collegiate Dictionary, which is registered with the U.S. Copyright Office. In another example, a prompt about a Britannica listicle, "10 Things You Need to Know About the Hamilton-Burr Duel," produced output that "identically reproduced the selection and ordering of the quotes" from the article, down to the snippets Britannica had chosen, and even noted that Britannica had fact-checked the material. The filing lists fifteen copyright registrations as exhibits. These allegations are organized into four copyright counts: direct infringement for training and RAG inputs, direct infringement for the outputs, and vicarious and contributory infringement against the various OpenAI entities. [4]
The fifth count is a trademark claim under the Lanham Act for false designation of origin and dilution. This is the more unusual theory, and it turns the problem of hallucination into a legal liability. When ChatGPT "generates made-up content or 'hallucinations' and falsely attributes them to Plaintiffs," the complaint argues, or when it quietly omits parts of an article and presents the incomplete version next to the Britannica or Merriam-Webster name, users are deceived into thinking the publishers stand behind the answer. The companies own federally registered "Britannica" and "Merriam-Webster" marks dating back to the 1980s. The harm, in their telling, is reputational: a wrong answer wearing a trusted brand erodes the very trust the brand was built on. [4][6]
The complaint also recounts a failed negotiation, which matters because OpenAI has signed content deals with other publishers, including News Corp and the Associated Press. According to the filing, Britannica and Merriam-Webster approached OpenAI in November 2024 to discuss licensing. The initial conversation went nowhere, "an OpenAI representative rebuffed Plaintiffs' licensing outreach," and OpenAI "never seriously pursued licensing Plaintiffs' content." The plaintiffs use this to preempt a fair-use defense: they argue ChatGPT creates "commercial substitutes" rather than transformative works, "adds no new expression, meaning, or message of their own," and could have paid for the content the way OpenAI paid others. [4]
The relief sought is broad. The plaintiffs ask for statutory and actual damages, restitution of OpenAI's profits, attorneys' fees, and, most consequentially, a permanent injunction barring the conduct described in the complaint. [4]
OpenAI rejected the claims. A spokesperson said the company's "models empower innovation, and are trained on publicly available data and grounded in fair use," and added that ChatGPT "helps enhance human creativity, advance scientific discovery and medical research, and enable hundreds of millions of people to improve their daily lives." That is the same defense OpenAI has raised across its copyright cases: that training on public web data is a transformative, lawful use. The complaint anticipates this argument and disputes it head-on, which sets up the central fight in the litigation. [1][7]
The suit is part of a wave of copyright litigation against AI developers. By the time it was filed, trackers counted roughly ninety such cases in the United States. The most closely watched is New York Times v. OpenAI, filed in late 2023, which raised the same core questions about training data and verbatim reproduction. Authors, music publishers, image libraries, and newspaper chains across the U.S. and Canada have brought parallel claims, and the broader doctrine is still unsettled: a 2025 ruling and settlement involving Anthropic suggested that training on copyrighted text can be transformative, yet acquiring that text through piracy is not protected. [2][3][8]
What distinguishes the Britannica case is the type of plaintiff. Newspapers sell yesterday's reporting; reference publishers sell definitions, dates, and curated facts, exactly the short, authoritative snippets a chatbot is most likely to surface and least likely to send a user elsewhere to read. That makes the "traffic cannibalization" theory more acute for an encyclopedia or dictionary than for almost any other kind of publisher. It is also not Britannica's first move in this fight. In September 2025, the same two companies sued Perplexity AI, the AI search startup, on closely related copyright and trademark grounds, accusing it of reproducing their content and attaching their brands to AI summaries. The OpenAI complaint reads as the second front in a deliberate campaign by reference publishers to define how answer engines may use their archives. [3][9]
The outcome could shape the economics of online reference work. If the courts accept that a verbatim or near-verbatim answer drawn from a publisher's archive, presented under that publisher's name, is infringement rather than fair use, AI companies may be pushed toward licensing deals with the sources they rely on for facts. If OpenAI's fair-use defense prevails, publishers that depend on search traffic will have to find another way to fund the human research behind their content. For now the case is in its earliest stages, with no ruling on the merits.