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

The legal battle over generative AI training data has reached a new milestone as Merriam-Webster and its parent company, Britannica, filed suit against OpenAI. This litigation represents a direct challenge to the fundamental mechanics of how large language models (LLMs) are developed and monetized. By alleging that ChatGPT "cannibalized" their web traffic, the plaintiffs are targeting the heart of the generative AI value proposition: the ability to synthesize and deliver information directly to users, bypassing the traditional need for source-site visits. This case moves beyond the simple question of whether training on copyrighted data is "fair use" and enters the territory of market displacement and unfair competition.

For decades, Merriam-Webster and Britannica have relied on a business model predicated on being the definitive authorities for definitions and factual queries. Their digital presence is sustained by advertising and subscription revenue driven by high-intent search traffic. The plaintiffs argue that OpenAI’s models were trained on their curated datasets to replicate their authoritative voice, effectively creating a derivative product that serves as a direct substitute. This argument is particularly potent in the context of reference materials, where the user's goal is a specific, factual answer rather than an immersive creative experience. When an AI provides that answer directly, the economic incentive to visit the original source vanishes, threatening the viability of legacy reference institutions.

“

The legal battle over generative AI training data has reached a new milestone as Merriam-Webster and its parent company, Britannica, filed suit against OpenAI.

OpenAI has historically defended its training practices under the doctrine of fair use, arguing that the transformative nature of AI training—creating a new, functional tool from existing data—does not infringe on the original works. However, the Merriam-Webster suit highlights a growing legal consensus that when an AI tool directly competes with the source material in its primary market, the fair use defense weakens. The plaintiffs are expected to argue that the training process was not transformative enough to justify the total displacement of their web traffic and revenue streams.