Federal Judge Affirms Right to Use Copyrighted Works for AI Training

Judge William Alsup ruled that Anthropic did not violate the Copyright Act by training its large language model (LLM) on legally acquired copyrighted books. This case arose from authors accusing Anthropic of “stealing” their works by using pirated versions. While Alsup acknowledged that Anthropic had downloaded millions of pirated books, he clarified that training LLMs with legally obtained materials is permissible for generating original content. He emphasized that expecting authors to compensate for each use of their work would be unreasonable, and affirmed AI’s right to transform content similarly to how humans do. Furthermore, Alsup argued that the use of copyrighted texts by AI does not displace demand for the authors’ original works. Although the Authors Guild plans to appeal, suggesting a reversal could lead to increased prices for AI tools, the current ruling is a significant affirmation of the transformative use of copyrighted material in AI training.

Source link