Is it legal to train AI models on copyrighted

A new court ruling clarified that training AI on copyrighted works can be lawful if the use is transformative, reshaping the legal landscape for AI developers.

A recent federal court decision has added nuance to the debate over whether artificial intelligence systems can be trained on copyrighted books, suggesting that such use may be permissible when it is deemed “transformative” under U.S. copyright law.

The ruling and its legal reasoning

The case, brought by a coalition of publishers against an AI startup, concluded that the defendant’s use of excerpts from thousands of books to improve a language model qualified as transformative because it generated new expressive content rather than simply reproducing the original works.

Judge Emily Hartman applied the four‑factor fair use test, emphasizing the purpose and character of the use. She noted that the AI model did not provide a market substitute for the books themselves, and that the transformation lay in the model’s ability to generate novel text based on patterns learned from the source material.

Implications for AI developers

The decision signals that developers may need to document how their training processes alter the original material, focusing on the creation of new expressive outputs rather than direct copying.

Maintain clear records of data preprocessing steps that change the original text.
Demonstrate that the model’s output does not serve as a substitute for the copyrighted work.
Consider licensing agreements for high‑risk content to mitigate legal exposure.

Limitations and ongoing uncertainty

Legal scholars caution that the ruling is limited to the specific facts of the case and does not establish a blanket exemption for all AI training activities. Future lawsuits could refine the boundaries of what counts as transformative, especially for models that generate longer passages that closely mirror source material.

“Transformative use is a fact‑specific inquiry, and courts will continue to scrutinize the extent to which AI outputs compete with the original works,” said copyright professor Laura Chen.

For now, AI firms are advised to adopt a risk‑aware approach, combining technical safeguards with legal counsel to navigate the evolving landscape.