Thaler v. Perlmutter: Why AI Can’t Hold a Copyright (and Why It Matters to You)

Every debate about AI and copyright eventually circles back to one deceptively simple question: can a machine be an author? Not “did a human use a machine” — but can the software itself hold the copyright to what it makes? For years that was a thought experiment. Then a scientist named Steven Thaler decided to test it for real, and turned it into the most important AI-authorship case in the country.

The result — Thaler v. Perlmutter — is the case people point to when they say “AI can’t be copyrighted.” That shorthand is close enough to be dangerous, because the ruling is much narrower than it sounds, and understanding the difference tells you exactly what you can and can’t do with your own AI-assisted work. Here’s the plain-English version.

A gold robotic hand stopped by a barrier just short of an empty ornate picture frame, symbolising a machine denied authorship of a work
The core question: can the machine itself be the author? The court said no.

The 30-second version

Steven Thaler built an AI system he named the Creativity Machine. He says it generated a piece of visual art on its own, titled A Recent Entrance to Paradise. When he applied to register the copyright, he did something unusual: he listed the Creativity Machine as the author — the actual creator — and himself only as the owner, on the theory that he owned the machine.

Track the cases yourself

Every current AI-copyright ruling — and the ones still being fought — lives in our AI Copyright Ruling Tracker. Filter by your creator type to see, in plain language, how each case affects your rights.

The US Copyright Office refused. Its position, unchanged for decades, is that a copyright needs a human author. Thaler asked it to reconsider; it refused again. So he sued the head of the Office, Register of Copyrights Shira Perlmutter. He lost in the district court, appealed, and on March 18, 2025 the US Court of Appeals for the DC Circuit affirmed: under US law, the author must be a human being, so the machine’s work couldn’t be registered as-is.

Why he set it up that way (this part is crucial)

Here’s the detail almost every headline skips, and it changes everything. Thaler didn’t argue “I made this with AI, so I’m the author.” He argued the opposite. He stipulated as a fact — locked it in as part of the case — that the machine created the work autonomously, with no meaningful human involvement, and he insisted the machine be named the sole author.

That was a deliberate strategy. Thaler wanted a test case establishing that an AI itself can be an author. But it also boxed the court in. Because the record said a machine made it alone, the judges only had to answer one question: can a non-human be the author of a copyright? They never had to wrestle with the messier, more common situation — a human using AI as a tool. Thaler chose the hardest possible ground to fight on.

A gold spotlight falling only on a human silhouette while a robot silhouette stays unlit, symbolising that copyright recognises only human authors
The law puts the spotlight on a human author, not the tool.

What the court actually held

The DC Circuit’s answer was clear and unanimous. Writing for the panel, Judge Patricia Millett held that the Copyright Act of 1976 “requires all eligible work to be authored in the first instance by a human being.” The word “author,” read through the whole statute, assumes a human: authors have life spans that set copyright terms, they can hold property and sign transfers, they have heirs, they have intentions. A machine has none of that. So the Creativity Machine could not be the author, and a work with no human author couldn’t be registered.

Notice what that reasoning does not say. It doesn’t say AI-made images are inherently worthless, or that using AI poisons your copyright, or that you can never protect anything you made with a generative tool. It says one specific thing: the law won’t recognize the machine as the author. The human question was left for another day — on purpose.

The doors the court deliberately left shut

Appeals courts are careful to decide only what’s in front of them, and this panel was explicit about the questions it was not answering. Three of them matter to you:

A spotlight on one small cube while larger question-marked cubes stay in shadow, symbolising a narrow ruling that left bigger questions open
The court answered one narrow question and left the bigger ones in the dark.
  • It didn’t rule on the Constitution. Thaler also argued that the Constitution requires (or at least allows) protection for machine authorship. The court set that aside — it decided the case purely on the statute.
  • It didn’t decide whether Thaler could own the output another way. He’d raised a theory that he should own the work as its maker or under work-made-for-hire principles, but because he had forfeited or reframed those arguments, the court didn’t resolve them.
  • It didn’t say how much human input is ‘enough.’ This is the big one. The whole live debate for working creators — where’s the line between “AI did it” and “a person made it with AI” — was simply not before the court, because Thaler had stipulated there was no human input at all.

In other words, the case that everyone cites as the final word on AI and copyright pointedly refused to answer the question most creators are actually asking.

So can you copyright anything you make with AI?

Yes — and this ruling doesn’t change that. The human-authorship requirement it confirmed has always coexisted with a simple reality: a tool doesn’t erase the human using it. Photographers use cameras. Musicians use synthesizers. Designers use software full of automated features. None of that stops the human from being the author, because the human is making the creative choices.

A human hand guiding a paintbrush that a robotic arm also holds, symbolising human creative control over an AI tool
The workable path: a human clearly steering the tool, not the tool acting alone.

The Copyright Office’s own guidance follows the same logic. It will register works that contain AI-generated elements as long as there’s meaningful human authorship — your selection, arrangement, editing, and creative control over the final result. What it won’t do is register raw output where a human just typed a prompt and accepted whatever came back, or — as in Thaler’s case — where the human insists a machine did everything. The protection covers your contribution, not the machine’s.

So the practical rule after Thaler is almost the reverse of the scary headline. Don’t disclaim your own authorship. If you shaped the work — chose it, refined it, combined it, edited it, arranged it — you’re the author, and you say so.

What this means for you

Whether you’re an illustrator, a musician, a YouTuber, or a small-business owner making your own graphics, here’s the honest read:

  • A machine can’t be your co-author — and it doesn’t need to be. Never list an AI as the author on a registration. You are the author; the AI is a tool you used.
  • Human creative control is the whole game. The more your choices, edits, and arrangement shape the final work, the stronger your claim. Pure one-prompt output with no human shaping is the weak spot.
  • Thaler is narrow. It killed one specific idea — that software itself can hold a copyright. It did not declare AI-assisted art unprotectable. Anyone telling you otherwise is over-reading it.
  • The line is still being drawn. Exactly how much human input is enough remains unsettled, and future cases and Copyright Office decisions will sharpen it. Watch that space.

What to do now

  • Document your creative process. Keep your prompts, drafts, iterations, edits, and the choices you made. If your authorship is ever questioned, that trail is your proof that a human — you — shaped the work.
  • Register the human-made work, and be straight about the AI. When you register, claim your original contributions and disclose AI-generated material honestly. Accurate applications hold up; misleading ones can be cancelled.
  • Add real human authorship on top of AI drafts. Edit, compose, arrange, combine, retouch. The value you add is exactly what copyright protects.
  • Don’t name the tool as the creator. Credit yourself. That single framing choice is, quite literally, the difference the court cared about.

Thaler v. Perlmutter gets remembered as the moment the courts said “AI can’t hold a copyright.” It did — but only because a scientist went out of his way to insist a machine, and only a machine, made his picture. For the rest of us, who use these tools the normal way, the ruling quietly reaffirms something reassuring: put a human’s creativity into the work, and the law still has your back.

This article is general information, not legal advice. AI and copyright law is moving fast and differs by country and state — for a specific situation, check the current law where you are or talk to a qualified attorney.

IPSurge helps creators and small businesses understand and protect their intellectual property — the ideas, names, and work that make them unique.


Sources & further reading:

Related Reading

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top