Two recent rulings from the Northern District of California affirmed the use of copyrighted books for training large language models (LLMs) as fair use. Both judges emphasized the transformative nature of the LLM training process and noted that the original works weren’t meaningfully available to the public. However, their analyses diverged on three key issues: the distinct purpose of LLM training, the implications of using pirated copies, and the reproduction’s impact on the market. Judge Alsup’s decision in Bartz v. Anthropic highlighted that training an LLM was fair use, as it did not negatively affect the authors’ market. Conversely, using pirated books for a central library was deemed unfair, as it directly displaced demand for the authors’ works. This contrast raises questions about how courts may interpret fair use concerning AI training moving forward. A follow-up analysis will compare these findings with Judge Chhabria’s decision in Kadrey v. Meta.
Source link
