AI getting spicy with a dash of IP: Encyclopædia Britannica and Merriam-Webster just sued OpenAI

Encyclopædia Britannica and Merriam-Webster just sued OpenAI in the Southern District of New York. This is another in a series of cases arriving at LLM’s doors. The complaint alleges that OpenAI scraped nearly 100,000 Britannica articles to train its models, continues to copy that content through RAG systems, and then gives answers that replace the original source, stripping the attribution and traffic links that fund Britannica’s business. That last part isn’t trivial: Traditionally, search engines sent traffic to publishers. The LLMs keep users from them.

The complaint also includes a Lanham Act claim: When ChatGPT, in this case, hallucinates and uses the Britannica or Merriam-Webster name on the output, that’s false designation of origin. The case lands in the already-active OpenAI MDL in SDNY, so expect coordination with other pending suits like those brought by the New York Times and Authors’ Guild.

OpenAI will be under pressure to settle. A jury looking at ChatGPT output alongside identical Britannica output will have an easy job.

How this differs from the earlier Google cases:

The 2015 Google decision (Authors Guild v. Google) turned on the fact that Google showed snippets and ultimately sent users to the sources. So Google was a pointer to the source, not a replacement for it. The Second Circuit bought the transformative use argument because Google’s product complemented the originals rather than acting as substitute.

OpenAI’s problem is the opposite. ChatGPT gives an answer in a closed loop. No click required, no visit to Britannica, no subscription sold.

That’s what makes this different and harder to defend on fair use grounds. Add in the RAG layer (real-time copying at the moment of each query, not just a one-time training scrape), along with the trademark/hallucination angle, and the complaint attacks OpenAI in ways that the Google courts never had to address.

The Google cases were about access to the underlying source; this case is about replacing that source without attribution or payment.