In the Tremblay v. OpenAI case, Magistrate Judge Illman granted the book authors’ request to compel OpenAI to turn over one of the datasets that it used: “English Colang Dataset.” It apparently was difficult to navigate under the data-inspection rules set for discovery by agreement of the parties, as reported by Aruni Soni of Bloomberg Law, who was at the hearing.
Of course, one of the ironic things about the datasets at issue in the respective cases is that a number of them were compiled by nonprofits or researchers–and are freely shared online. OpenAI, however, asserts that it has a “proprietary dataset.”