The Algorithm's Muse: AI in Authorship and Publishing

Updated Oct 2026: this editorial has been rewritten and now cites its sources.

Generative AI has moved from novelty to everyday tool for some writers, and the publishing trade is still working out what that shift means for authorship. The questions are concrete. Who owns a text that a machine helped produce? What must a self-publisher disclose? Can AI companies train on published books without permission? Courts, the US Copyright Office, retailers and authors' organizations have each offered partial answers. This editorial places those answers side by side and draws out what they mean for authors deciding how, or whether, to use AI.

Authorship requires a human

The legal starting point in the United States is that copyright protects human creativity. In 2023 a federal judge upheld the Copyright Office's refusal to register an artwork that its creator said was made autonomously by an AI system, writing that human authorship is a "bedrock requirement of copyright". The case was brought by Stephen Thaler over an image titled "A Recent Entrance to Paradise." Ars Technica reported that Judge Beryl Howell of the US District Court for the District of Columbia framed the sole issue as whether a work generated entirely by an artificial system, absent human involvement, should be eligible for copyright, and ruled that it should not.

The Copyright Office has since laid out its position in detail. Its 2025 report on copyrightability concludes that copyright does not extend to purely AI-generated material and that, given current technology, prompts alone do not give a user enough control over the output to make that user its author. Human contributions, such as selecting and arranging material or making creative modifications, can still be protected.

For an author, the practical effect is that the parts of a book a person writes or meaningfully shapes are protectable, while the parts generated wholesale by a machine may not be.

What retailers require

Retailers have added rules of their own. In September 2023 the Guardian reported that Amazon would require self-publishers to declare whether content is AI-generated, defining AI-generated content as text, images or translations created by an AI-based tool, while not requiring disclosure of AI-assisted work. Around the same time Amazon limited KDP accounts to creating three new titles a day, a step the Guardian linked to concerns about AI-generated books.

Those rules turn the distinction between generating and assisting into a matter of compliance, not only of conscience.

The risks are not purely regulatory. In August 2023 Amazon removed five books listed under the name of publishing writer Jane Friedman after she complained that they were falsely attributed to her. Friedman said she believed the books were written by AI. The case shows that machine-generated text can be used against real authors as well as by them.

The training question

The most contested issue is whether AI developers may train their models on published books without a license. On 20 September 2023 the Authors Guild and 17 authors, among them John Grisham, Jodi Picoult, David Baldacci and George R.R. Martin, filed a class action against OpenAI in the Southern District of New York on behalf of fiction writers whose works were used to train GPT. The Guild later amended its complaint to name Microsoft as a defendant. Publishers and authors have also sued Meta, alleging that the company torrented protected books from pirate sites such as LibGen and Anna's Archive to develop its Llama models. These are allegations made in the complaint.

On 1 September 2026 the US Department of Justice filed a statement of interest in the consolidated OpenAI litigation in New York supporting the position that training can be fair use. A statement of interest is not a ruling, and the court will decide the question. Until it does, and until any appeals run their course, the law on training remains unsettled.

What authors' organizations advise

The Authors Guild has published AI best practices for authors, which urge authors to preserve their own voice and to disclose substantial AI use to publishers and readers. The Guild's guidance also spells out two practical consequences. AI-generated text included in a work submitted for copyright registration must be disclosed and disclaimed in the application. And because book contracts contain a warranty that the manuscript is original to the author, including AI-generated text in a final manuscript may violate that warranty. That guidance lines up with the legal position above: the more of a book that reflects the author's own work, the clearer both its ownership and its integrity.

A framework for authors

Taken together, these sources suggest a simple framework:

• Authors should write the core themselves. Human authorship is what copyright protects, according to the courts and the Copyright Office.
• Know the retailer's rules. On Amazon, AI-generated content must be declared at publication.
• Keep records. Notes on how a book was made help answer questions about authorship later on.
• Check contracts. The Authors Guild warns that AI-generated text may breach a contract's originality warranty.
• Follow the litigation. The training cases will shape how authors' work can be used.

The muse and the machine

AI tools can help with tasks around writing, from brainstorming to checking consistency. The sources above nonetheless draw a clear line: the law protects what a human creates, retailers ask for disclosure when a machine does the creating, and the use of authors' books for training is being decided in court. Authors who keep their own voice at the center of their work stay on the right side of the first two lines; the third will be settled by the courts.