Even as its lawyers defended piracy before federal judges, OpenAI began to sign deals with major international media companies for the use of their copyrighted content as training data. Because now we can point to a thriving market for legally licensed AI training data (see Exhibit A above, courtesy of Ezra Eeman) and an actual price paid for the use of that training data. In early January, documents in the Kadrey v. Meta lawsuit, a leading copyright infringement case against Meta and its Llama AI model, revealed that members of Meta’s AI team were clearly aware that they were using (to quote their own words) “pirated material” to train their model.