New York Times Accuses OpenAI and Microsoft of Copyright Infringement
Summary
On December 27, 2023, the New York Times (NYT) filed a complaint in the Southern District of New York (SDNY) against Microsoft and OpenAI, alleging that unauthorized copies of NYT works were used to train GPT models; that “memorized” copies are encoded in the parameters of the LLMs as either copies or derivative works; and related allegations that the models provide unauthorized reproductions of NYT works. Responsive filings are due in the coming months, however, OpenAI published a statement refuting the allegations as “without merit.” In its statement, OpenAI emphasizes its collaboration with news organizations, indicates that training is fair use, and identifies that “memorization” is an uncommon bug that primarily relies on misuse of the model (e.g., engineering prompts to cause malfunction).
