Is it legal to train AI models on copyrighted books? It’s complicated
techcrunch ·
The article explains that training AI on copyrighted books raises complex legal questions that are not yet settled. A recent ruling ordered Anthropic to pay a $1.5 billion settlement, but the judge actually found its AI training lawful, penalizing only the use of pirated books from illegal shadow libraries. The key legal doctrine is fair use, which depends on factors like whether the use is transformative and whether it competes with the original work. Courts have been inconsistent, as seen in Thomson Reuters v. Ross Intelligence, where copying content to build a competing AI-based legal platform was not considered fair use. Additionally, purely AI-generated works may not be copyrightable, which raises further unresolved questions about authorship and AI assistance. Overall, the copyright law, unchanged since 1976, has not caught up with AI technology, and many cases remain in litigation.
techcrunch