Law and The Machine

The Ghost in the Copyright Machine: Why Stephen Thaler and His AI Lost at the Supreme Court

March 20, 202613:28Law and The Machine

This episode explores the Supreme Court's definitive rejection of copyright protection for fully autonomous AI-generated content, stemming from computer scientist Stephen Thaler's test case. Listeners will learn about the "bedrock requirement" of human authorship in US copyright law, supported by statutory interpretation and historical precedents like the "monkey selfie" case. The discussion highlights the consistent legal stance that non-human entities cannot be authors or inventors under current US intellectual property law.

Key Takeaways

Detailed Report

The U.S. Supreme Court has issued a definitive ruling, effectively closing the door on copyright protection for content generated solely by artificial intelligence. This decision reinforces a long-standing legal principle: human authorship is a fundamental requirement for copyright in the United States.

The Thaler Test Case

This legal clarity stems from a deliberate challenge by computer scientist Stephen Thaler. Thaler sought copyright for an artwork titled "A Recent Entrance to Paradise," explicitly listing his "Creativity Machine" as the author and himself only as the owner of the machine. He adamantly maintained that he provided no creative input, prompts, edits, or guidance, aiming to create a pure test case for autonomous AI authorship.

A Consistent Legal Defeat

Thaler's quest met with an unbroken chain of rejections. His application was first denied by the U.S. Copyright Office, a decision upheld by the District Court, the D.C. Circuit, and finally, the Supreme Court, which declined to hear the case, effectively cementing the lower courts' rulings. This consistent judicial stance underscores a deep-rooted legal consensus.

The "Bedrock Requirement" of Human Authorship

While the Copyright Act of 1976 doesn't explicitly state "must be human," courts have consistently interpreted its language to imply human authorship. Key arguments include:

Statutory Interpretation

The Act's provisions, such as measuring copyright protection by the "life of the author plus 70 years" and referencing an author's "widow or widower," "children," or "grandchildren" inheriting rights, are uniquely human concepts. The legal framework simply doesn't accommodate non-human entities with these attributes.

Historical Precedent

Courts leaned on historical cases like *Burrow-Giles Lithographic Co. v. Sarony* (1884), which established that copyright protects the "intellectual conception" of the author – the human mind's creative choices, not merely mechanical reproduction.

Modern Precedent

The "monkey selfie" case (*Naruto v. Slater*), where a macaque took selfies, further solidified this principle. The Ninth Circuit ruled that animals lack statutory standing under the Copyright Act, reinforcing that non-humans cannot be authors.

Parallel in Patent Law: The DABUS Case

Thaler pursued a similar battle in patent law, attempting to have another of his AI systems, DABUS (Device for the Autonomous Bootstrapping of Unified Sentience), recognized as an inventor. The Federal Circuit, handling patent appeals, ruled that the U.S. Patent Act requires an "inventor" to be a "natural person." This outcome, and its reasoning, mirrored the copyright decisions, establishing a unified front across U.S. intellectual property law.

Internationally, a broad consensus is forming. The UK Supreme Court and the European Patent Office have also dismissed Thaler's appeals in the DABUS case, emphasizing that AI systems lack legal personality and cannot own or transfer property rights. While South Africa granted a patent listing DABUS as an inventor, its non-substantive examination system makes it an outlier that doesn't break the global agreement.

Implications for the Generative AI Industry

This ruling has significant implications for the burgeoning generative AI industry:

  • Public Domain Content: Purely machine-generated content, devoid of sufficient human creative input, is likely to fall into the public domain, meaning it cannot be exclusively owned or licensed.
  • Focus on Human-in-the-Loop: Businesses must now strategically focus on, and crucially document, the human contribution to AI-assisted works. The protectable intellectual property lies in the human element.
  • U.S. Copyright Office Guidance: The USCO distinguishes between AI as a tool assisting human creativity (which can lead to protectable works) and AI as a replacement for it (which does not).
  • Prompts Insufficient: Merely providing text prompts is generally considered insufficient for authorship, as prompts are viewed as unprotectable ideas, with the AI retaining control over the expressive execution.
  • Disclosure Requirement: The USCO now requires applicants to disclose the use of AI-generated material in copyright applications and describe their human creative contributions.

The Next Frontier: Defining Human Input

While Thaler's case provided a clear "no" to purely autonomous AI authorship, the more complex and commercially relevant legal battles are just beginning. The case of *Allen v. Perlmutter* is a key one to watch.

Jason Allen, who won a digital art competition with his Midjourney-created piece "Théâtre D'opéra Spatial," is suing the Copyright Office after it refused to register his work. Allen argues that his intensive creative process, involving over 624 distinct text prompts, subsequent Photoshop edits, and upscaling, constitutes significant human creative control, analogous to a photographer using a camera.

The "Threshold of Originality"

The core question in *Allen* and future cases will be: How much human input, iterative prompting, and post-processing is "enough" to qualify for authorship? This involves grappling with difficult questions:

  • Is there a quantitative measure for creative input?
  • Where does the creativity truly lie: in the initial prompt, iterative refinement, or final selection and editing?
  • How does the "black box" nature of some AI models, where output is not fully predictable, affect claims of creative control?

Thaler's case established that the "ghost in the copyright machine" must be human. The next chapter will be about determining just how substantial that human ghost needs to be to claim ownership of what the machine produces, defining the nuanced boundaries of creative control in an AI-driven world.

Show Notes

Works Referenced

Glossary

Sources / References

Full Transcript

HostSo, the Supreme Court just officially slammed the door shut on AI-generated content getting copyright protection. Not in some nuanced, careful way, but with a pretty definitive "no."
ExpertNot just shut, they basically nailed it shut with a one-line order, at least for fully autonomous AI. This wasn't even a debate for them. It cements a bright-line rule: if a human isn't behind the creative wheel, it’s not getting U.S. copyright.
HostAnd this all stems from a legal quest by a computer scientist named Stephen Thaler, who explicitly tried to force the issue, right? He basically said, "My AI created this, I had zero creative input, now give it a copyright."
ExpertExactly. Dr. Thaler made it impossible for the courts to sidestep the core question. He applied for copyright for an artwork called "A Recent Entrance to Paradise," listing his "Creativity Machine" as the author. His name was only on it as the *owner* of the machine. He was intentionally, strategically, *adamant* that he provided no prompts, no edits, no guidance. He wanted a pure test case. And he got his answer.
HostHe sure did. This decision didn't come out of nowhere, either. The report highlights a string of losses for Thaler, starting with the U.S. Copyright Office itself, then the District Court, the D.C. Circuit, and finally, the Supreme Court refusing to even hear it.
ExpertIt's been a consistent, unbroken chain. The Copyright Office initially rejected it, citing a long-standing requirement for human authorship. Then the District Court, in a ruling from Judge Beryl Howell, called human authorship a "bedrock requirement." The D.C. Circuit upheld that, saying the Copyright Act requires a human being. And now the Supreme Court just nodded along by denying certiorari. No surprises here, for anyone who's been following IP law.
HostSo, what *is* this bedrock requirement? Because the Copyright Act of 1976 doesn't explicitly say, "must be human," does it?
ExpertIt doesn't use those exact words, no. But the courts, particularly the D.C. Circuit, did a deep dive into statutory interpretation. They argued that if you look at the text *as a whole*, humanity is a necessary condition. Think about it: copyright protection is measured by the "life of the author plus 70 years." Machines don't have legally cognizable lifespans, right? How would you apply that?
HostGood point. A machine doesn't have a birth certificate or a death certificate in that sense. And I doubt it has a will.
ExpertPrecisely. The Act also talks about an author's "widow or widower," "children," or "grandchildren" inheriting rights or terminating transfers. These are uniquely human concepts. The idea of a machine having heirs or a spouse is, frankly, absurd in the legal context. Even the "work-made-for-hire" doctrine, which Thaler tried to use, was dismissed because that doctrine *presumes* the work was first created by a human employee or contractor.
HostSo it's not about what the law *doesn't* say, but what it *does* say, and how that implies a human being at its core.
ExpertAbsolutely. It’s about the underlying assumptions baked into the law. The courts also leaned heavily on historical precedent, going all the way back to *Burrow-Giles Lithographic Co. v. Sarony* in 1884. That case established that copyright protects the "intellectual conception" of the author, essentially the human mind's creative choices. It wasn't about the mechanics of taking a photograph, but the photographer's artistic input – the posing, the lighting, the composition. That's the link.
HostWait, *Burrow-Giles*? That's the famous Oscar Wilde photograph case, right? So the courts are essentially saying a Midjourney image is closer to a mechanical reproduction than a photographer's artistic choices, *if* there's no human guiding hand.
ExpertExactly. If the AI is making all the creative choices autonomously, then it's essentially a sophisticated machine just outputting something. And the report highlights another, more recent, and frankly, funnier precedent: the "monkey selfie" case.
HostAh, yes, Naruto the macaque. That was a wild one.
ExpertWild, but legally very relevant here. In *Naruto v. Slater*, a crested macaque took selfies, and PETA sued on the monkey's behalf, arguing Naruto was the author. The Ninth Circuit ultimately ruled that animals lack statutory standing under the Copyright Act, pointing to those same textual clues we just discussed about "children" and "widows."
HostSo, if a monkey can't be an author, neither can a "Creativity Machine." It sounds like the courts are pretty consistent on this "non-human cannot be author" principle.
ExpertThey really are. And it wasn’t just copyright where Thaler fought this battle. The report details a parallel legal battle where he tried to get another of his AI systems, DABUS – Device for the Autonomous Bootstrapping of Unified Sentience – recognized as an inventor on patent applications.
HostDABUS! I remember that.
ExpertThat's right. And the outcome, and the reasoning, were virtually identical. The Federal Circuit, which handles patent appeals, ruled that the U.S. Patent Act requires an "inventor" to be an "individual," which they clarified unambiguously means a natural person. Again, Thaler tried to argue policy – that allowing AI inventorship would "promote the Progress of Science" – but the court just said, "Nope, the statute's clear."
HostSo it's a unified front then, across both copyright and patent law in the U.S. And if I recall, this wasn't just a U.S. thing.
ExpertNot at all. The report makes it clear there's an international consensus forming. The UK Supreme Court, just in late 2023, also dismissed Thaler's appeal in the DABUS case, ruling that an "inventor" must be a natural person. The European Patent Office came to the same conclusion earlier. They all emphasize that AI systems lack legal personality and can't own or transfer property rights.
HostSo, most major IP regimes globally are in agreement. Except for South Africa, which apparently granted a patent listing DABUS as an inventor.
ExpertYes, but the report notes that South Africa has a non-substantive patent examination system, meaning they don't scrutinize inventorship claims in the same way. So it's a bit of an outlier, and doesn't really break the global consensus. The takeaway is pretty clear: AI is a sophisticated tool, but it's not a legal creator in the eyes of the law, at least for now.
HostThis has huge implications for the generative AI industry, doesn't it? If purely AI-generated content can't be copyrighted, what does that mean for businesses built around it?
ExpertIt’s a massive deal. The immediate consequence is that purely machine-generated content, without sufficient human creative input, is likely in the public domain. This includes a vast and growing amount of images, text, music, and code. So, if your business model relies on owning and licensing *purely* autonomous AI outputs, you have a problem.
HostIt's like... you've got this incredible generative engine, but everything it spits out is free for anyone to use.
ExpertPrecisely. It forces a strategic shift. Companies now have to focus on and, crucially, *document* the "human-in-the-loop." The value, and the protectable IP, lies in the human contribution. The U.S. Copyright Office has been pretty proactive on this, putting out guidance. They distinguish between AI as a tool to assist human creativity – which is fine and protectable – and AI as a *replacement* for it, which is not.
HostSo, if I use Midjourney to generate an image, but then I take that image into Photoshop and spend hours editing it, arranging elements, adding my own filters and textures... *that's* protectable?
ExpertYes, exactly. The report indicates that the USCO says that creatively selecting, arranging, or significantly modifying AI-generated material *can* result in a copyrightable work. But, and this is key, the copyright would only cover your human-authored contributions, not the underlying AI-generated components.
HostSo you can't claim ownership of the *entire* image if the AI did the initial heavy lifting. And what about prompts? Because a lot of people see crafting elaborate prompts as a creative act in itself.
ExpertThe USCO is pretty clear on that too. In a January 2025 report, they stated that merely providing a text prompt is likely insufficient for authorship. They view prompts as "unprotectable ideas" or instructions, while the AI system itself retains control over the expressive execution of those ideas. It's like telling a sculptor, "make me a unicorn." The idea is yours, but the sculptor's execution is where the creativity, and thus the copyright, lies.
HostThat's a significant distinction. So, no matter how clever my prompt, it's not my "intellectual conception" in the legal sense.
ExpertNot in the way the law currently defines it for copyright. And the USCO even requires disclosure now: if you use AI-generated material, you have to state that in your copyright application and describe your human creative contributions. The burden of proof is now firmly on the human to demonstrate their authorship.
HostSo, Thaler's case was a "blunt instrument," as the report puts it, asking a simple yes-or-no question. But what happens when the lines are blurry? This feels like it's just the beginning.
ExpertYou're absolutely right. That's the next legal frontier, and it's far more complex and commercially relevant. The report highlights *Allen v. Perlmutter* as the case to watch. This involves Jason Allen, who won a digital art competition with his piece "Théâtre D'opéra Spatial," created with Midjourney. The Copyright Office refused to register it, and Allen sued.
HostWhat makes *Allen* so different?
ExpertAllen claims he engaged in an incredibly intensive creative process. He used *over 624 distinct text prompts* to refine the image. Then he took that output and did corrective edits in Adobe Photoshop and upscaled it with Gigapixel AI. His argument is that this iterative, deeply involved process is analogous to a photographer using a camera. He provided the "intellectual conception," and Midjourney was merely the tool that executed his vision.
Host624 prompts? That's a serious amount of iteration. It sounds like he's trying to show he had significant creative control.
ExpertExactly. But the Copyright Office counters that prompts are still just ideas, and the expressive output is ultimately generated by the AI, not Allen. They argue that his time and effort are irrelevant; what matters is creative *control* over the final expression. Though, interestingly, the office did note he could have sought a limited registration for just his Photoshop edits, but he's seeking copyright for the *entire* work.
HostSo, the core question in *Allen* becomes: How much human input, how much iterative prompting, how much post-processing, is "enough" to qualify for authorship?
ExpertThat’s it. It’s the "threshold of originality" for AI-assisted works. The courts will have to grapple with really difficult questions. Is there a quantitative measure? Does 600 prompts confer authorship where 6 might not? Where does the creativity really lie – in the initial prompt, the iterative refinement, or the final selection and editing? And what about the "black box" nature of some AI models, where you can't fully predict the output? Does that affect your claim of creative control?
HostIt feels like we're heading into a very messy, very subjective legal battle. Will we need expert witnesses just to explain how AI models work, and how much "control" a human actually has?
ExpertIt's highly likely. The *Thaler* decision established that the ghost in the copyright machine must be human. The next chapter will be about determining just how substantial that ghost needs to be to claim ownership of what the machine produces.
HostSo, the Supreme Court said, "No copyright for purely autonomous AI." That's clear. But the real fight is just beginning as we try to define that human boundary.
ExpertExactly. The initial legal certainty provides a temporary framework, but it also highlights the complexities that were deliberately avoided in Thaler's test case. The future of creative ownership is going to be defined by these gray areas.
HostSo what does "creative control" actually mean in an age of increasingly sophisticated black-box AI tools? And how will these legal battles shape public perception of the value, or lack thereof, of AI-generated content?