In a landmark ruling, Judge William Alsup sided with Anthropic in a copyright case, declaring that training AI models on legally purchased books constitutes fair use. This decision specific…
In a landmark ruling, Judge William Alsup sided with Anthropic in a copyright case, declaring that training AI models on legally purchased books constitutes fair use. This decision specifically addresses the practice of Anthropic buying physical books, digitizing them, and using the digital copies to train its Claude AI models.
The judge emphasized that digitizing purchased books and utilizing them for training LLMs is transformative enough to be considered fair use, drawing parallels to the educational process of teaching writing skills. This victory marks a significant win for the AI industry, setting a precedent for similar copyright cases in the future.
However, the ruling also highlights that Anthropic is not entirely in the clear. Judge Alsup determined that the company must face a separate trial concerning the alleged pirating of "millions" of books from the internet. This part of the case focuses on Anthropic's actions of storing a vast library of pirated books, even if some were not directly used for training.
The judge expressed skepticism that downloading source copies from pirate sites, when legal acquisition was possible, was reasonably necessary for fair use. The lawsuit was initiated by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who claimed that Anthropic trained its AI models on pirated material.
The upcoming trial will determine the extent of damages resulting from the alleged copyright infringement related to the pirated books. This aspect of the case underscores the complexities of copyright law in the age of AI and the ongoing debate over the boundaries of fair use in AI training.
Anthropic has expressed satisfaction with the court's recognition of the transformative nature of using works to train LLMs. A spokesperson emphasized that Anthropic's models are designed to create something different, fostering creativity and scientific progress, rather than replicating or supplanting existing works.
This statement reflects the company's position that its AI training practices align with the core purpose of copyright law, which is to encourage innovation and the creation of new works.