On March 13, 2026, Encyclopaedia Britannica and its dictionary subsidiary Merriam-Webster filed a federal lawsuit against OpenAI, alleging that the company copied nearly 100,000 of their encyclopedia entries and dictionary definitions — without permission, license, or payment — to train the models behind ChatGPT. The case is Encyclopaedia Britannica, Inc. v. OpenAI, Inc., No. 1:26-cv-02097, filed in the Southern District of New York under the Copyright Act (17 U.S.C. § 101 et seq.) and the Lanham Act, and assigned to Judge Sidney H. Stein. Within days it drew coverage from Reuters, Fortune, CNET, Law.com, TechCrunch, and the IP trade press — one of the more heavily covered filings on the IP Feed, and a rare one where a reference publisher with two and a half centuries of history squares off against the best-known name in generative AI.
What sets this suit apart from the broader wave of AI-training disputes is that it does not rest on copyright alone. Britannica pairs a familiar copyright theory — that its catalog was ingested wholesale to build a commercial product — with a trademark and false-endorsement theory that few other plaintiffs have pressed: that ChatGPT puts words in Britannica's mouth. That second front turns the AI industry's most notorious weakness, hallucination, into a legal liability, and it is the part of this complaint worth watching most closely.
What the lawsuit says
The copyright claim is by now familiar. The plaintiffs allege that OpenAI scraped their reference works — professionally researched, edited, and fact-checked over decades — and used them to teach ChatGPT to answer questions, without a license. They go further than pure input, though: the complaint alleges that ChatGPT's outputs reproduce their material in full or near-verbatim form, and that a system trained to hand users a finished answer diverts the very readers who would otherwise visit Britannica.com or Merriam-Webster.com — draining the subscription and advertising revenue that pays for the work. The complaint also targets retrieval-augmented generation, the feature that lets ChatGPT pull live information from the web, alleging it scans and surfaces Britannica's content in real time.
The trademark claim is the novel part. Britannica alleges two distinct harms. First, that ChatGPT presents invented, sometimes inaccurate content under Britannica's and Merriam-Webster's names — attaching a trusted brand to text the publishers never wrote. Second, that ChatGPT displays partial reproductions of their material in ways that imply the companies endorse, or are the source of, the answer. Both, the complaint argues, mislead users under the Lanham Act by trading on brands built over generations.
The defendant has not yet answered these allegations on the merits, and no court has ruled on them. OpenAI has said publicly that its models "are trained on publicly available data and grounded in fair use." What follows is the legal question the filing raises — not a prediction of how it comes out.
The hard part: fair use — and a twist
Every one of these cases runs into the same doctrine, and it usually decides them: fair use. OpenAI argues, as AI developers generally do, that training on copyrighted works is transformative — that the model learns statistical patterns rather than storing and replaying the text, serving a different purpose than the originals.
Britannica attacks that on the fourth fair-use factor, the effect on the market. Its argument is unusually concrete: unlike a novel or a news feature, a reference entry exists to answer a specific question, and a chatbot that answers that question from the entry is not serving a different market — it is capturing the same one, upstream. When the output substitutes for a visit to the source, the market harm is not hypothetical; it is the business model.
Then there is the twist the copyright analysis cannot reach. Fair use is a defense to copyright infringement; it is not a defense to trademark false-endorsement. If Britannica can show that ChatGPT attaches its name to content it did not write, that claim lives or dies on likelihood of confusion, not on whether the training was transformative. That is why the hallucination angle matters beyond this case: it offers rights holders a theory that sidesteps fair use entirely. Courts have not settled how the Lanham Act applies to AI outputs that invoke a brand, and this is one of the first high-profile cases to test it.
Why it matters beyond one lawsuit
For anyone who owns a catalog of creative work, this litigation is about more than encyclopedias. The traffic-substitution argument — that an AI answer replaces the trip to the source — applies to any business whose value lies in being the authoritative version of something: a film library, a music catalog, a stock of licensed footage. And the false-endorsement theory raises a question every brand owner will eventually face: what happens when a model speaks in your name?
There is a quieter lesson for rights holders too. Cases like this turn on the ability to prove exactly what you created, when, and who owns it — the same provable chain of authorship and ownership that underpins the value of any IP asset. When the fight is whether your work was copied, whether its market was harmed, and whether your name was misused, a clean record of what you own is not paperwork; it is leverage. For independent creators without a reference publisher's litigation budget, that record is often the only leverage there is.
We'll track this docket as it develops. The first real signals will come as OpenAI files its response and the shape of its fair-use defense emerges, as the court tests whether the trademark claims survive alongside the copyright ones, and as discovery probes how the training corpus was assembled and how often ChatGPT reproduces reference content. Follow the filings, counsel, and coverage on the case page.
This post is editorial commentary on public court filings and news coverage, not legal advice. The allegations described are unproven, and the defendant has not yet responded on the merits. Characterizations of the complaint reflect press coverage of the filing, not any court finding.