The AI and Copyright Dilemma: Can machines be authors? 

In a decision released on March 18^th, 2025, the United States Court of Appeals for the District of Columbia Circuit ruled that non-human machines cannot hold copyright over the works they create.¹ The decision hinges on a fundamental principle of the US Copyright Act of 1976, which establishes that only human beings can be recognized as the authors of their work.  

The Creative Machine  

The case centers around Stephen Thaler, a computer scientist who developed a generative AI system that he named the Creative Machine (“CM”). In 2019, Thaler submitted a copyright application for an artwork titled A Recent Entrance to Paradise, listing CM as the author. However, the application was promptly denied, with the reason given that “a human being did not create the work”.² Thaler’s subsequent argument that he should be considered the owner of the work under the work-for-hire doctrine was also rejected, as there was no contract between him and CM.  

After receiving the rejection, Thaler requested reconsideration from the Review Board, arguing that the requirement for human authorship was unconstitutional and outdated, given the more modern practices of AI product generation. He further contended that judicial opinions needed to evolve to accommodate computer-generated creations.  

The District Court Upholds the Review Board’s Decision 

Following a second rejection from the Review Board, Thaler took his case to the United States District Court for the District of Columbia, seeking summary judgment. In addition to his earlier points, Thaler argued that his “instructions and direction” to CM should make him eligible for copyright.³ However, the court upheld the Review Board’s decision, affirming that “human authorship is a bedrock requirement of copyright”.⁴

The Court of Appeal’s Clarity on Human Authorship 

In its March 2025 decision, the Court of Appeal thoroughly examined of the use of the term “author” under the US Copyright Act. Notably, the court points out that the Act does not explicitly define “author”, but through traditional statutory interpretation, “author” was held to refer exclusively to human beings, not machines.⁵ The court highlighted that machines are merely tools used by humans in the creative process, not the creators themselves.  

The court referred to seven key statutory provisions to support its ruling, each of which makes sense only if “author” is understood to mean a human being: 

1. Ownership: Copyright is fundamentally a property right granted to authors. Since a machine cannot own property, it cannot be considered an author under the law. 

2. Duration: Copyright is linked to the author’s lifespan, which is inherently tied to human life. As machines do not have lifespans, this provision does not apply to AI-generated works. 

3. Inheritance: The Act’s inheritance provision states that when an author dies, their “termination interest is owned, and may be exercised by their ‘widow or widower’ or ‘surviving children or grandchildren’” none of which can be had by machines.  

4. Signature: the Act requires a signature to transfer copyright ownership, something a machine is unable to provide.  

5. Nationality or Domicile: The Act protects unpublished works regardless of the author’s “nationality or domicile”. However, machines have neither, further underscoring their inability to be considered authors under the Act.  

6. Intention: Machines do not have the legal intention required to create original works in the same way that human authors do.  

7. Machines as Tools: The Act makes consistent references to machines as tools, rather than authors. Machines are repeatedly described as devices or processes that assist their human creators, not as creators themselves.  

Collectively, these provisions confirm that authorship under the Copyright Act is inherently tied to humanity. As the court aptly put it, “the human-authorship requirement, in short, eliminates the need to pound a square peg into a textual round hole by attributing unprecedented and mismatched meanings to common words in the Copyright Act.”⁶

The court was careful to note that it was not asserting any specific statutory provision as an essential requirement for claiming authorship. Instead, the court’s central argument is that, when read in its entirety, the Copyright Act implies that humanity is a necessary condition for authorship under federal law. 

In response to Thaler’s argument that he could be the author under the work-for-hire doctrine, the court underscored the importance of the word “considered” in the statutory language. This word ensures that copyright and authorship rights for works created by humans are automatically transferred to the person who hired the creator. The court pointed out that, had Congress intended for non-humans to be included, the statute would have clearly stated that those who hire creators “are the author,” not that they are merely “considered the author.”⁷

A Focus on Law, Not Policy 

The court also emphasized that its role was to interpret and apply the law as it is written, rather than making decisions based on what might be considered “good policy.” In response to Thaler’s argument that these judicial opinions are outdated, the court clarified that it was not suggesting AI cannot play a role in creative processes. Instead, it maintained that under current US copyright law, AI cannot be recognized as the sole author of a work 

While this case raised important questions about the intersection of technology and copyright law, the court’s decision reaffirms a key principle: authorship is a uniquely human right under US copyright law. The ruling does not negate the role that AI can play in the creative process; rather, it reinforces the idea that only human beings can hold authorship rights. 

A Canadian Perspective 

In Canada, copyright law hinges on the concept of “originality”, which is analogous but not identical to the US concept of “creativity”. To qualify for protection in Canada, a work must be original, meaning it is the product of the author’s “skill and judgment”, created independently and without copying from other sources. It does not, however, require a work to be completely novel or groundbreaking – what matters most is that it reflects the genuine intellectual effort of the author.  

The landmark Canadian case of CCH Canadian Ltd v Law Society of Upper Canada established these principles, noting that the exercise of skill and judgment “will necessarily involve intellectual effort” and “must not be so trivial that it could be characterized as . . . purely mechanical.”⁸ This standard, while not as clear in its application to AI as the American counterpart, offers a solid peg to rest an argument on to suggest that human authorship is an essential component of originality in Canadian copyright law.  Further, many of the factors pointed out by the US court (like references to ownership, lifetime-based duration, nationality, and assignments) also appear in the Canadian Copyright Act. 

Canadian case law in this area remains relatively sparse. In 2021, the Canadian Intellectual Property Office (CIPO) registered a copyright for a painting titled Suryast, listing two co-authors – one human, the other an AI machine. This marked a significant step in challenging the traditional notion of authorship under the Copyright Act. However, as the court pointed out in Thaler, co-authorship can be distinguished from sole authorship.⁹

In the same year as the Suryast registration, Canada also released a public consultation paper to explore potential amendments to the copyright framework. Its aim was to address the increasing influence of AI in the creative process and explore how they could adapt to a rapidly evolving technological landscape.¹⁰ Topics of discussion included reconsidering the definition of an author, granting author rights to the human creators behind AI, and establishing a unique set of rights for AI-generated works. No major changes have been implemented yet.   

More recently, the Samuelson-Glushko Canadian Internet Policy and Public Interest Clinic (“CIPPIC”) filed a Notice of Application to the Federal Court, seeking to expunge the copyright ownership granted to the AI machine in Suryast.  

As a public interest legal clinic, CIPPIC argues that granting copyright authorship to an AI-generated image raises critical questions about authorship, originality, and the role of AI in creative industries. Their claim centers on several issues, including the image’s lack of originality due to insufficient skill and judgment, and the fact that AI, as a non-legal entity, cannot hold authorship under Canadian copyright law. That application is presently pending, with the parties considering filing expert evidence on issues including the operation of generative AI in the creation of art. The matter is expected to be heard later this year.   

Unanswered Questions: The Complexities of AI, Ownership, and Moral Rights 

The possibility of a machine being granted ownership of a work raises many interesting – and tricky – questions. The court in Thaler addressed the issue of ownership and assignment, touching on whether an AI could ever assign a work it “created”. The conclusion seems clear: without legal capacity or the ability to sign a contract, it’s highly unlikely that an AI machine could assign ownership.  

However, one aspect Thaler did not address is the concept of moral rights. In Canadian copyright law, authors are granted not only copyright over their works, but also moral rights. These rights include the right of integrity, the right of anonymity, and the right of association. Moral rights are inherently tied to the creator’s identity and cannot be assigned or transferred – though they can be waived. Moral rights are powerful in that they defend the emotional and reputational well-being of creators. This raises a crucial question: can an AI machine truly possess the moral framework to hold these rights, or are they fundamentally tied to human identity?  

As AI continues to evolve and make its mark on creative industries, debates over authorship and ownership are bound to intensify. But as the Thaler decision and Canada’s public consultation paper show, meaningful change in copyright law will likely require more than just a fresh interpretation of the statutes – it may need a full legislative overhaul. And so, the AI authorship saga marches on. 

Footnotes

1. Stephen Thaler v Shira Perlmutter, No. 23-5233 (DC Cir 2025), March 18, 2025 [Thaler]. ↩︎
2. Thaler at 6. ↩︎
3. Thaler at 8. ↩︎
4. Thaler at 8. ↩︎
5. Thaler at 9. ↩︎
6. Thaler at 12. ↩︎
7. Thaler at 17. ↩︎
8. CCH Canadian Ltd. v Law Society of Upper Canada, 2004 SCC 13 at para 16. ↩︎
9. Thaler at 19. ↩︎
10. Government of Canada, A Consultation on a Modern Copyright Framework for Artificial Intelligence and the Internet of Things, (Ottawa: Innovation, Science and Economic Development Canada, 2021) at section 2.2.1. ↩︎