US Federal Judge Rules AI Can Train on Copyrighted Books

The case, Bartz v. Anthropic PBC, was filed by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson. They claimed that Anthropic, an AI company, used their copyrighted books without permission to train its AI model, Claude. The authors argued this use was a violation of their exclusive copyright rights, which protect their works from unauthorized reproduction and distribution.

Judge Alsup’s decision was based on the principle of "transformative use." In simple terms, he found that when AI trains on copyrighted works, it doesn’t just copy the material but learns from it, extracting patterns and structures to create new, original outputs. This concept of transformation is important in copyright law and has been used in other cases, such as Andy Warhol Foundation v. Lynn Goldsmith, where the Supreme Court held that creative works that add new meaning or expression can qualify as fair use.

The judge also considered the purpose of AI training. He acknowledged that AI research and development have a public benefit, advancing technology in ways that could eventually benefit society. Anthropic's use of copyrighted books, he argued, contributed to this broader goal of innovation, making it more likely to be deemed fair use.

When it came to the amount of the works used, Judge Alsup noted that while large datasets are necessary for effective AI training, Anthropic didn't copy the authors' books in full. Instead, the AI process focuses on learning from various sources without directly replicating any single work. This weighed in favor of the AI company’s argument.

However, Judge Alsup wasn’t willing to give AI companies a free pass. While he ruled in favor of Anthropic in this case, he stressed that the fair use defense does not apply to all uses of copyrighted works in AI training. Each case will still need to be evaluated based on its specific facts. He made it clear that AI companies can’t assume they can use copyrighted content without consideration of the context and purpose of that use.

This ruling has significant implications for the AI industry, which often relies on vast amounts of data, including copyrighted works, to train models. The decision provides a legal framework for AI companies to use such content under fair use, but also introduces some caution. It signals that while AI development can benefit from copyrighted materials, companies need to be mindful of the legal and ethical boundaries.

Many experts believe this ruling will influence other ongoing legal battles involving AI and copyright, particularly those involving major players like OpenAI and Meta. As more companies develop and deploy AI technologies, courts will likely continue to shape the balance between innovation and intellectual property rights.

Judge Alsup's ruling in Bartz v. Anthropic offers a more straightforward legal path for AI companies using copyrighted works in their training models. While it supports innovation in AI, it also serves as a reminder that fair use is not a blanket defense, and companies must navigate these legal waters carefully. The case is a win for AI developers, but it’s also a signal that the intersection of copyright law and technology is an area that will continue to evolve.