Merriam-Webster and Britannica Sue OpenAI for 'Cannibalizing' Dictionary Data

The legal battleground for generative AI has expanded from the realms of creative literature and visual arts into the foundational world of reference data. Merriam-Webster and its parent company, Britannica, filed a lawsuit on March 17, 2026, against OpenAI, claiming that the AI giant’s flagship product, ChatGPT, was built using vast quantities of their copyrighted dictionary and encyclopedia entries without authorization. This case represents a significant escalation in the ongoing conflict between content owners and AI developers, shifting the focus to the highly structured, factual data that gives Large Language Models (LLMs) their linguistic precision and semantic depth.

At the heart of the complaint is the allegation that OpenAI’s training process involved scraping millions of definitions, etymologies, and usage examples that Merriam-Webster and Britannica have spent decades—and in some cases, centuries—curating. While facts themselves are generally not copyrightable under U.S. law, the specific expression, arrangement, and comprehensive compilation of those facts in a dictionary are protected. The plaintiffs argue that ChatGPT does not merely provide information but replicates the unique voice and structural expertise of their reference works, effectively creating a derivative product that competes directly with the original sources.

“

At the heart of the complaint is the allegation that OpenAI’s training process involved scraping millions of definitions, etymologies, and usage examples that Merriam-Webster and Britannica have spent decades—and in some cases, centuries—curating.

The economic argument presented by the publishers is particularly compelling: the cannibalization of web traffic. For decades, the business model for digital dictionaries has relied on search engine traffic leading users to their websites, where ad revenue and premium subscriptions sustain the costly work of lexicography. By providing instant, high-quality definitions within its own interface, ChatGPT bypasses the need for users to visit external reference sites. Merriam-Webster and Britannica contend that this is not a case of transformative fair use, but rather a parasitic relationship where the AI uses the publishers' own data to render their primary distribution channels obsolete.

This lawsuit follows a pattern of litigation from other content-heavy industries, including the New York Times and various authors' guilds. However, the Merriam-Webster case is unique because it targets the very building blocks of language that AI models require to function. If the courts side with the publishers, it could force a radical shift in how AI companies source their training data. We may see the emergence of a mandatory licensing regime for reference data, similar to how music streaming services pay royalties to labels and artists. For OpenAI, which has recently sought to strike licensing deals with news organizations like Axel Springer and the Associated Press, this lawsuit suggests that the fair use defense is becoming increasingly difficult to maintain as a blanket strategy.