Anthropic won the case: judge allowed AI training on books without authors’ consent

Anthropic company won the case against book authors and American judge William Alsup issued a truly revolutionary decision. Recognizing as lawful the use of published books for training artificial intelligence models. Without authors’ permission. This is the first case where a court supported the position of technology companies about the applicability of fair use doctrine to training neural networks on copyrighted materials.

The decision became a serious blow to authors, artists and publishers who filed dozens of lawsuits against OpenAI, Meta, Midjourney, Google and other companies. Although the verdict doesn’t guarantee that other judges will follow Alsup’s example, it creates a legal foundation for future decisions in favor of technology companies.

Notably, the court still appointed a separate hearing on the issue of Anthropic’s so-called “central library”. The company downloaded 7 million copyrighted books from pirate sites to create its database, which is clearly illegal. The judge noted, I quote: “That Anthropic later bought a copy of a previously stolen book doesn’t free the company from responsibility for theft. But may affect the size of the fine”.

Interesting is that the basis for court proceedings becomes the interpretation of fair use doctrine. A special section of copyright law that hasn’t been updated since 1976. Long before the appearance of the internet and the concept of generative artificial intelligence.