Merriam-Webster and Britannica Sue OpenAI Over AI Training 'Theft'

The legal landscape for generative AI has shifted into a critical new phase as Merriam-Webster and Britannica, two of the world’s most venerable reference publishers, filed a joint lawsuit against OpenAI. The litigation alleges that the San Francisco-based AI giant "stole" proprietary material to train its ChatGPT models, effectively creating a competing product that directly undermines the plaintiffs' primary revenue streams. This case represents a significant escalation in the ongoing conflict between traditional content owners and the builders of large language models (LLMs), moving beyond creative prose into the realm of foundational, authoritative data.

At the heart of the complaint is the concept of "traffic cannibalization." Merriam-Webster and Britannica argue that by ingesting their entire corpora of definitions, etymologies, and encyclopedic entries, OpenAI has built a system that provides users with the same utility without the need to visit the original websites. For reference publishers who rely heavily on ad revenue driven by search engine traffic, this represents an existential threat. If a user asks ChatGPT for a complex definition or a historical summary, and the AI provides it using data derived from these sources, the original publisher loses the opportunity to monetize that inquiry. This "zero-click" reality is a growing concern for the entire digital publishing industry, which has already been reeling from changes in search engine algorithms.

“

The legal landscape for generative AI has shifted into a critical new phase as Merriam-Webster and Britannica, two of the world’s most venerable reference publishers, filed a joint lawsuit against OpenAI.

This lawsuit follows a precedent set by The New York Times and various authors' guilds, but it carries unique weight because of the nature of the content involved. Dictionaries and encyclopedias are often viewed as collections of facts, which have traditionally been harder to copyright than creative fiction. However, the specific arrangement, selection, and expression of those facts—the "editorial craft"—is legally protected. OpenAI is likely to lean on the "fair use" defense, arguing that its training process creates a transformative new tool rather than a derivative work. Yet, the plaintiffs contend that ChatGPT’s output is often a near-verbatim or highly derivative reflection of their proprietary databases, which they have spent centuries refining and verifying.