Britannica and Merriam-Webster sue OpenAI over ChatGPT training data

In short: Encyclopedia Britannica and Merriam-Webster are suing OpenAI, saying it used their reference content to train ChatGPT without permission.

What happened

Encyclopedia Britannica and Merriam-Webster filed a lawsuit against OpenAI in Manhattan federal court on March 13, 2026. They claim OpenAI copied and used almost 100,000 encyclopedia and dictionary items to help train ChatGPT.

The publishers say ChatGPT can return answers that are “near-verbatim,” meaning very close to the original wording. They argue this competes with their own sites and could reduce visits and subscriptions.

Britannica also points to something called retrieval-augmented generation, or RAG. In simple terms, this is when a chatbot looks things up from a database while it is writing an answer (like checking notes mid-sentence). Britannica says OpenAI’s use of its articles in this process breaks copyright rules. The lawsuit also claims trademark problems, including that ChatGPT sometimes cites Britannica as a source even when the answer is wrong.

Britannica is asking for money damages and a permanent court order that would stop the alleged copying. OpenAI said its models are trained on publicly available data and that this is protected by “fair use,” but it did not address the specific claims.

Why it matters

This case is part of a bigger fight over whether AI companies can learn from copyrighted writing without a license. Courts have not fully settled this question yet, so the result could affect what information chatbots can use and whether publishers get paid.

Source: The Verge AI