Encyclopedia Britannica has filed suit against OpenAI in Manhattan federal court, joining the growing list of publishers taking legal action over unauthorized use of copyrighted content in AI training. The case is particularly notable given Britannica's 250-year history as a knowledge authority and the potential precedent for reference works.
Britannica has been around since before copyright law existed in its current form. Now they're suing over whether an AI can learn from their content. This case could define whether "reading" content to train AI counts as fair use or theft.
The lawsuit, filed March 16, alleges OpenAI used Britannica's articles and reference materials to train its large language models without permission or compensation. This follows similar suits from The New York Times, authors groups, and other publishers who argue AI companies are profiting from copyrighted works without paying for them.
The stakes here are existential for reference publishing. Britannica's business model is creating and maintaining authoritative, fact-checked content. If AI companies can ingest that content for free, train models on it, and then compete with Britannica by answering the same questions, what's the incentive to create quality reference materials?
OpenAI's argument—deployed in other cases—is that training AI models on published content constitutes transformative fair use. They're not copying articles to republish them. They're learning patterns from them to create new content. That's fundamentally different from piracy.
Publishers counter that the distinction is meaningless if the result competes with the original. If users ask ChatGPT questions instead of consulting Britannica, the harm to Britannica's business is the same whether OpenAI copied articles or just learned from them.
Courts will eventually decide whether AI training constitutes fair use. But the legal framework wasn't designed for this use case. Fair use doctrine considers factors like purpose, nature, amount used, and market effect. AI training is commercial (purpose), uses entire works (amount), and potentially competes with the original (market effect). That cuts against fair use.
But AI training is also transformative, creating something new rather than substituting for the original. And if every use that potentially competes is prohibited, fair use becomes meaningless.
