AI training is fair use, Wired cofounder Kevin Kelly argues in OpenAI copyright case
- Published on Oct 21, 2026 at 02:37 PM IST
Wired founding executive editor Kevin Kelly has supported OpenAI and other defendants in a US copyright case. He argues that using copyrighted books to train AI models should count as fair use. He says this use helps encourage the creativity that copyright law is meant to support.
In a proposed brief supporting the court as a non-party, dated October 16, Kelly asked the court to rule in favour of the defendants on the issue of AI training. The brief was submitted as an exhibit in the combined OpenAI copyright case in the Southern District of New York.
Kelly, who writes both fiction and nonfiction, argues that training large language models creates a new kind of technology rather than a replacement for the books used to build them. His main point is that copyright protects an author's original way of expressing something, but does not give authors exclusive control over the ideas, facts and knowledge in their works.
“Copyright is not a shield from disruptive technology,” the brief says.
The brief compares AI training with how people learn by reading and gaining experience. It explains that AI models study large amounts of text to find patterns and connections. They can then use what they learn to help create new material.
Kelly argues that writers have long taken inspiration from earlier works without necessarily breaking copyright law. The brief gives two examples: Shakespeare's Romeo and Juliet, which drew on earlier stories, and Jane Smiley's A Thousand Acres, a modern retelling of King Lear. The argument is that using shared ideas and themes is different from copying the original words and other protected expression.
The brief also claims that AI models do not store works in their original expressive form. This claim is part of Kelly's argument for fair use. It has not been established as a court finding in the proposed brief.
One important business issue in the filing is whether competition from books made with AI should count as harm under copyright law. Kelly describes the plaintiffs' position as seeking protection from new works that do not infringe copyright but compete with their books. He argues that this kind of competition is different from giving readers copies that replace the original books.
To show how AI can help writers, the brief refers to self-published author Aaron Black, who reportedly published four books in ten months while writing the prose himself. According to the account cited in the filing, Black used AI to help with research, keep details consistent across his manuscripts and find visual references. He remained in control of the creative work.
Kelly's wider argument is that these tools can give individual writers help that was once more commonly available from teams of researchers, editors and assistants.
The brief also points to AI uses in medical research, mathematics and criminal investigations. These examples support its argument that limits on AI training could slow scientific and creative progress in many areas. Kelly asks the court to consider these benefits when deciding whether the training qualifies as fair use.
Kelly accepts that AI raises serious questions about what the technology can do and how it affects society. However, he argues that copyright law should not be used to settle every concern about AI. In his view, allowing the training described in the case would better serve the basic purpose of copyright.