The dictionary takes OpenAI to court

For generations, Merriam-Webster and Encyclopedia Britannica stood as the ultimate gatekeepers of reliable knowledge. Today, people just ask ChatGPT. But the publishers of those gold-standard references are drawing a line in the sand.

Encyclopedia Britannica and its subsidiary Merriam-Webster filed a lawsuit against OpenAI, accusing the AI giant of massive copyright infringement.

Britannica, which owns the rights to nearly 100,000 online articles, claims OpenAI scraped and fed that vast catalog into its model training pipelines without asking for permission or paying a dime.

Scraped for training, squeezed by RAG

The lawsuit extends far beyond initial training data. Britannica argues OpenAI infringes copyright by generating responses that reproduce its content verbatim, either in full or in part.

It also takes direct aim at OpenAI's retrieval-augmented generation (RAG) system, where ChatGPT crawls and pulls Britannica articles to answer real-time web queries. According to the complaint, this setup starves web publishers of revenue by serving substitute answers that compete head-to-head with original sources.

Britannica added another twist by invoking the Lanham Act, a federal trademark statute. The publisher claims ChatGPT fabricates hallucinations and falsely attributes that misinformation to Britannica or Merriam-Webster, undermining hard-won reputations.

This legal barrage follows similar fights across media. OpenAI is already facing lawsuits from The New York Times, Ziff Davis, and dozens of North American newspapers. Britannica itself has an ongoing case against Perplexity.

Courts are still sorting out fair use in AI training. While a federal judge previously ruled that Anthropic’s training process was legally transformative, the same judge faulted the company for downloading pirated books, which helped lead to a $1.5 billion settlement with writers. OpenAI has not commented on the new filing.

Why it matters

This is one of the sharpest legal assaults OpenAI has faced, because Britannica is attacking where AI search hurts most.

Arguing over pre-training copyright often gets bogged down in fair use debates. But targeting RAG workflows and false attribution under the Lanham Act strikes at the actual product experience. When ChatGPT acts as an answer engine that pulls authoritative definitions, strips the web traffic, and occasionally hallucinates under Merriam-Webster's brand name, the commercial harm looks concrete.

For users, services like ChatGPT feel like magic reference tools today. But if AI answer engines hollow out the business models of original encyclopedias and dictionaries, the web's foundational baseline of verified human knowledge starts to decay.

We think OpenAI's free-for-all scraping era is hitting an immovable wall. With Anthropic's $1.5 billion precedent looming, licensing payouts will likely become an inevitable cost of doing business.