You’ve spent six months writing a novel. You outlined the plot yourself, built the characters from scratch, and wrote every chapter with your own fingers on the keyboard. You also used ChatGPT to brainstorm a few scene ideas, asked Claude to suggest a better opening line for chapter three, and ran your dialogue through Sudowrite to punch it up. The book is unmistakably yours.
But when you sit down to register the copyright, a question nags at you: is it still yours, legally?
What AI Copyright Means
AI copyright isn’t a single rule. It’s a tangle of legal questions that emerged when machines started producing things that look a lot like creative work. Who owns a novel when AI helped write it? Can you copyright a cover image that Midjourney generated? If an AI model trained on your published book without permission, does that violate your rights?
These questions fall into two broad categories. The first is about outputs: can AI-generated content receive copyright protection, and if so, who owns it? The second is about inputs: is it legal for AI companies to train their models on copyrighted works? Both matter enormously for authors, and the answers are still being written (by humans, for now).
A Very Brief Legal History
Copyright law was built on a simple assumption: creative works come from people. The U.S. Constitution grants Congress the power to protect “Authors” and their “Writings,” and for centuries, nobody needed to ask whether the author was human. Then AI got creative, and the assumption needed testing.
The first real test came from Dr. Stephen Thaler, a computer scientist who filed a copyright registration in 2019 for an image his AI system generated. He listed the AI as the sole author. The Copyright Office said no. Thaler sued, and in August 2023, Judge Beryl Howell ruled that human authorship is a “bedrock requirement of copyright.” The D.C. Circuit Court of Appeals agreed unanimously in March 2025, though the judges made an important addition: this ruling does not prevent copyright for works where a human author “uses or is assisted by” AI. The door closed on AI-as-author but stayed open for AI-as-tool.
The case that drew the clearest line for creators came earlier, in February 2023, when the Copyright Office reviewed Zarya of the Dawn, a graphic novel by Kris Kashtanova. The book combined Kashtanova’s original writing with images generated by Midjourney. The Office’s decision was surgical: the human-written text and the creative arrangement of the whole book were protected, but the individual AI-generated images were not, because Kashtanova didn’t have enough control over what Midjourney produced. It was the first time the Office spelled out that AI-assisted works exist on a spectrum, and that copyright protection follows the human contribution, not the AI contribution.
The Rules as They Stand
In January 2025, the U.S. Copyright Office released its most comprehensive guidance yet. The key principles are straightforward, even if the edges are still fuzzy.
US copyright requires human authorship. Purely AI-generated expression is not protected by copyright. Human authors and organizations can own copyright in qualifying human-authored work; authorship and ownership are different questions.
Using AI does not disqualify the human contribution. Original human-written passages, creative arrangements, and sufficiently creative revisions may qualify for protection. The protection does not automatically extend to the AI-generated material within the same work.
Prompts alone aren’t enough. Writing even the most detailed, carefully crafted prompt doesn’t make you the author of the output. The Copyright Office compared it to giving instructions to a commissioned artist: you directed the work, but you didn’t create the expression. To claim copyright, you need to contribute original creative expression to the final work, not just instructions.
Registration requires disclosure of more than minimal AI-generated material. The US Copyright Office’s guidance requires applicants to identify more than a de minimis amount of AI-generated content and describe the human contribution. AI-generated portions are excluded from the claim.
The Other Side: Training on Your Work
There’s a second copyright question that hits even closer to home. The large language models behind tools like ChatGPT and Claude were trained on enormous datasets that included published books, often without the authors’ knowledge or consent. A dataset called Books3, containing nearly 200,000 books scraped from pirate sites, was used to train several major models.
Authors pushed back. In 2023, the Authors Guild filed a class-action lawsuit against OpenAI, joined by George R.R. Martin, John Grisham, Jodi Picoult, and dozens of other writers. Similar suits were filed against Meta, Anthropic, and others. By late 2025, more than 70 copyright lawsuits had been filed against AI companies. The core question in all of them: does using copyrighted training data to build an AI model constitute infringement, or is it fair use? Courts are still deciding, and their answers will shape the economics of AI for years to come.
What This Means for Your Writing Life
For authors using AI, the relevant question is which expressive elements they actually created. Selecting a tool or directing its output does not by itself establish copyright in everything it produces.
A few smart habits will protect you. Keep records of your process: save your drafts, your prompts, and your revisions. When you register your copyright, be upfront about where AI was involved and describe your own contributions clearly. Most importantly, make the work yours. Use AI to brainstorm, to get unstuck, to explore possibilities, but put your voice, your judgment, and your creative fingerprint on the final product.
The US Copyright Office assesses the human contribution case by case. Records of your drafts and revisions can help explain that contribution, but they do not guarantee protection for AI-generated passages.