Note: Indie Author Magazine, Indie Author Training, and our other sister sites maintain a neutral stance on artificial intelligence. In our coverage, we seek to provide an objective, informative take on the technology, and we encourage authors to make decisions about whether to use AI platforms based on their personal values and what’s best for their business.
This article is for informational purposes only and does not constitute legal advice. Consult a qualified attorney for guidance on your specific situation.
Only a few months after ChatGPT emerged in November 2022, as more users were discovering generative AI platforms and exploring their capabilities, authors and other creatives were already posing questions about AI and copyright. Would AI-generated content be copyrightable? What about works created by a mix of human and AI input? Would copyright protection extend to works being used to train AI models?
In IAM’s April 2023 special issue on artificial intelligence, guest author Ash Roberts explored where copyright law stood at the time—but with court cases still pending and without updated rulings from the Copyright Office, many of the answers authors wanted were still hypothetical.
Ash Roberts
Three years later, appellate court decisions, a three-part report from the U.S. Copyright Office, more than fifty active lawsuits against AI companies, and the $1.5 billion Bartz v. Anthropic settlement have provided authors with many of the answers they were looking for. Some questions remain unanswered, and court precedents and the developments in generative technology have raised new ones. But the legal landscape for creatives today is much clearer than in 2023, offering definitions and distinctions that matter to all authors.
Whether you choose to use AI or simply seek to understand the protections that exist for your work in its midst, here’s where things stand in 2026.
AI Cannot Be an Author. That’s Now Settled (Mostly).
Roberts’s original article detailed one of the standout cases taking place at the time of its publication. In Thaler v. Perlmutter, Stephen Thaler argued he should own the copyright to artwork his AI system, the Creative Machine, generated autonomously. In his original copyright registration application, which was denied, he listed the Creative Machine as the author of the work but claimed a “fruit of the tree” doctrine—if he owned the machine, he owned what it produced. At the time, that case was still pending.
It’s not pending anymore. In March 2025, the DC Circuit Court of Appeals unanimously affirmed the lower court’s ruling: The Copyright Act requires all eligible work to be authored by a human being, and the Creativity Machine doesn’t qualify. Multiple provisions within the Act make clear that authors must be humans, not machines, according to the court’s ruling.
Thaler filed a petition asking the US Supreme Court to take the case in October 2025, so there’s a slim chance the case will get revisited at the highest level. But for now, the legal standard is that AI acting on its own cannot hold copyright.
In the wake of Thaler’s case, it’s important to note that the court has yet to rule definitively on where the line is drawn for copyright protection when a human uses AI as a tool and contributes significant creative input along the way. Although this is likely the question many indie authors care about, the answer is yet to be decided.
The Copyright Office Got Specific.
Between 2024 and 2025, the US Copyright Office released a three-part report on AI and copyright that filled in a lot of gaps. The first part, published in July 2024, covered digital replicas, or “the use of digital technology to realistically replicate an individual’s voice or appearance,” according to the report. The subsequent parts, which were published in January 2025 and May 2025, covered the copyrightability of AI and the programs’ use of copyrighted works in training new AI models, respectively—topics many authors have wondered about. Although the Copyright Office’s report doesn’t have the force of law, it signals where the government thinks the line should be drawn.
On whether your AI-assisted work is copyrightable (Part 2, January 2025): The Copyright Office confirmed that prompts alone—even extraordinarily detailed, multi-step, heavily engineered prompts—do not make you the author of whatever the AI produces. Legally, prompts function as instructions, not creative expression. The prompt author only tells the machine what they want; the machine decides how to execute it.
This doesn’t mean AI-assisted work can never be copyrighted. The report laid out scenarios where copyright protection likely applies: If you input your own copyrighted writing and it’s perceptible in the output, for example, you’re the author of at least that portion. If you creatively select, arrange, or modify AI-generated content, those human contributions can be copyrightable. And hundreds of works incorporating some AI-generated material have been registered successfully since the 2023 guidance, so this isn’t a blanket ban on using AI in your process.
The practical spectrum looks something like this: If you’re brainstorming with AI, using it to check grammar, or letting it help you refine prose you wrote, your copyright is intact. If you’re generating passages with AI and then substantially rewriting them, this is also likely protectable, though a court would need to evaluate eligibility case by case. But if you’re prompting AI to write your entire book and hitting publish, the work is not copyrightable under current standards.
On whether AI companies can train on copyrighted works (Part 3, May 2025): The Copyright Office released a 108-page report examining fair use in AI training and concluded that certain uses of copyrighted material for training cannot be defended as fair use. Where AI-generated outputs are substantially similar to the training data that went in, the Office sees a strong argument for infringement.
Three Courts, Three Cases, Three Different Answers
Outside of the Copyright Office’s reports, several landmark court cases have laid the groundwork for future rulings around copyright and AI. Legal analysts have started calling the first batch of fair use rulings “the Fair Use Triangle” because the three decisions point in different directions depending on the circumstances.
Thomson Reuters v. ROSS Intelligence (February 2025)—The AI company lost. ROSS Intelligence used Westlaw’s copyrighted legal headnotes to train a competing AI-powered legal search engine. The court rejected the fair use defense for two reasons: The copyrighted material was subject to full protection, and the resulting AI product competed directly against the original. When your AI tool is designed to replace the thing it was trained on, fair use is a hard sell.
Bartz v. Anthropic (June 2025)—The AI company won on fair use for purchased books, though it settled with authors whose books were pirated. A group of authors sued Anthropic—the company behind Claude—for using their books to train its AI model. The judge found fair use for legally purchased books, calling training large language models “among the most transformative” uses the court expected to see. The case continued on the separate question of whether Anthropic’s use of pirated copies obtained through BitTorrent for the same purpose was also considered fair use—a distinction the court treated very differently. The case ultimately settled for $1.5 billion, though following a fairness hearing in April, final approval remains pending as of this publication.
Additionally, a set of authors who opted out of the initial Bartz v. Anthropic settlement are suing the AI company for copyright infringement in a separate case. The authors argue they “did not discover, and could not have reasonably discovered, that their specific works were included in Anthropic’s private, internal database” until the direct notice to class members was sent by the original plaintiffs. The case is still in early stages, so no rulings have been made yet.
Kadrey et al. v. Meta Platforms, Inc. (June 2025) — The AI company won, but the court left a roadmap for future plaintiffs. Authors including Sarah Silverman and Richard Kadrey sued Meta for training its collection of AI models, called LLaMA, on pirated copies of their books. Meta prevailed because the plaintiffs didn’t present sufficient evidence that Meta’s use harmed the market for their original works. But the court’s detailed reasoning argued the authors simply made the wrong argument in their motion and told future plaintiffs how to build a stronger case, particularly around the indirect ways AI-generated content can substitute for and devalue original creative work.
The scorecard so far: Three judges have ruled on fair use in AI training. Two of those have sided with AI companies, but one sided with the copyright holder. Yet with more than fifty active lawsuits regarding AI and copyright protections, filed by everyone from The New York Times to Disney to the Authors Guild, there’s a lot more to come.
The Case Indie Authors Should Watch
Roberts’s 2023 article mentioned Jason Allen, the artist who won a Colorado State Fair award with AI-generated artwork after spending eighty hours and nine hundred iterations creating his piece. He’s now suing the Copyright Office in the case of Allen v. Perlmutter after they denied his copyright registration.
Allen argues his case is fundamentally different from Thaler’s. Where Thaler listed the AI as the sole author and himself as merely the owner, Allen is arguing that he exercised significant creative control through more than six hundred prompts and extensive iteration—that the AI was his tool, not his replacement. His filing specifically distinguishes his level of involvement from the autonomous generation at issue in Thaler.
This case tests a scenario many authors navigate daily: using AI with substantial human direction, iteration, and creative decision-making. The outcome will go a long way toward defining where “assisted by AI” ends and “generated by AI” begins. As of June 2026, the case is still pending in federal court in Colorado.
What Hasn’t Changed
Some fundamentals from 2023 remain solid. Copyright still attaches automatically the moment you create something; registration with the Copyright Office is only required if you need to enforce it in court, and a DMCA takedown notice handles most situations indie authors encounter. You still don’t own the copyright to most elements on your book cover—the stock photos, typography, and design elements belong to their respective creators unless rights are explicitly transferred to you. Courts and the Copyright Office maintain that non-human creators can’t hold copyright, citing precedent in Naruto v. Slater that says Congress has not specifically extended copyright law to animals. And the terms of service for whatever AI tool you’re using still govern your right to use its output—most platforms transfer usage rights to you, but it’s always worth checking.
Other countries continue to handle the gray area of protecting creative works in the wake of artificial intelligence differently. Some jurisdictions assign copyright of AI-generated material to the programmer or the person who arranged for the work to be created. If you’re publishing internationally, the rules in each market may vary.
What to Do
If you use AI in your creative process: Use it as a tool within your workflow, not as a substitute for your creativity. Document your human contributions—keep drafts, revision notes, and records that show your creative decision-making. Substantially modify any AI-generated content before it goes into your final manuscript. And when registering work with the Copyright Office, disclose your AI involvement honestly, and describe your human contributions clearly. The Copyright Office has registered hundreds of works with AI components since 2023; transparency works in your favor.
If you’re concerned about your books being used to train AI: You’re in good company; dozens of lawsuits are pursuing exactly this issue. The legal landscape is still developing, with major decisions expected through 2026 and beyond. Monitor opt-out mechanisms offered by AI platforms where they’re available, and keep an eye on the Authors Guild and similar organizations that are actively litigating on behalf of creators.
“Substantial human intervention” is still the standard for copyright, just as it was in 2023. The difference is that we now have court rulings, Copyright Office reports, and billion-dollar settlements that offer additional definition. The law is catching up to the technology—but just like AI, it’s always evolving and likely won’t stay still for long.
This article is a companion to "Humans, Monkeys, and the Muddled Mess of Who Owns Copyright Over AI" published in April 2023. It reflects the legal landscape as of early 2026. Copyright law around AI continues to develop rapidly; check back for updates as new rulings are issued.
