Is training AI on copyrighted books legal? The answer is complicated
US court rulings on whether AI companies can legally train models on copyrighted works remain inconsistent, as decades-old copyright law struggles to address modern AI. Legal experts say outcomes often hinge on whether the resulting product directly competes with the original work's market.
Major AI chatbots like ChatGPT, Gemini, and Claude are trained on vast troves of text — including books and articles — often without authors' knowledge or consent. While this might seem clearly illegal, the legal reality is far more nuanced.
Last year, Judge William Alsup ordered Anthropic to pay $1.5 billion to a group of authors whose works were used in AI training. Crucially, though, the judge ruled that the training itself was lawful — Anthropic was penalized specifically for sourcing the books from illegal shadow libraries, not for the act of training on them. Alsup compared an AI model absorbing text to a writer studying literature to create something new rather than replicating it.
Attorney Cathy Gellis argues the ruling actually favors AI companies, since it treats AI training more like "reading" a work than "copying" it — and copyright law is built around copying, not mere use or consumption of a work.
Outdated law, shifting precedent
US copyright law hasn't been updated since 1976, forcing courts to apply old standards to new technology. Attorney Jason Henderson notes that court reasoning has been inconsistent, but a pattern is emerging: judges tend to rule against AI companies when the resulting product directly competes with the source material's market. That was the case in Thomson Reuters v. Ross Intelligence, where a court found that using Reuters' content to build a competing legal research platform wasn't fair use.
A separate complication involves copyright over AI-generated content itself. In Thaler v. Perlmutter, courts ruled that a fully AI-generated work isn't copyrightable, raising further questions about how to measure the degree of AI involvement in any given piece of work.
With most major AI companies still facing ongoing litigation, a definitive legal framework remains far off. In the meantime, early rulings continue to shape industry behavior, even though future court decisions could still overturn them.
