AI-Generated Works and the Future of Intellectual Property in the US and EU.
With a plethora of new AI tools available for creative people, the principles of intellectual property and its protection get challenged more and more. Can we consider tools like ChatGPT or Midjourney as assistive tools, or creators in their own right? What about the output of such tools? Can it be considered as the intellectual property of the human person who initiated and/or controlled the creative process, or is the output “owned by” the AI and thus uncopyrightable?
These are some of the questions that have sparked a lively debate among artists, lawyers, scholars, and policymakers in recent years. The rapid development and adoption of generative AI tools, which can produce text, images, music, and other forms of content based on large datasets of existing works, pose new challenges and opportunities for the creative industries. On the one hand, these tools can enhance human creativity and innovation by providing new sources of inspiration, feedback, and collaboration. On the other hand, they can also raise legal, ethical, and social issues regarding the ownership, authorship, originality, quality, and value of the generated works. In this blog post, we will first define AI-generated and AI-assisted works, and then explore how the U.S. and EU approaches to copyright protection address these complex issues, highlighting their similarities and differences.
Note! This blog post is not a legal advice. The sole purpose is to promote the discussion about copyright law and practices in the age of AI.
Defining AI-generated and AI-assisted works
Before diving into the U.S. and EU approaches to copyright protection, let's clarify the distinction between AI-generated and AI-assisted works. According to the scientific literature, such as the paper "Artificial Intelligence and Intellectual Property Law" by Anke Moerland, AI-generated works refer to creative outputs produced solely by an artificial intelligence system without human intervention. They are also known as "computer-generated works" or "machine-generated works." On the other hand, AI-assisted works refer to creative outputs where a human creator collaborates with an AI system during the creation process. In these cases, the AI system serves as a tool or creative assistant under the human creator's control.
Moerland argues that purely AI-generated works, which do not require human intervention, are still currently out of our reach, as they would require artificial general intelligence that has not yet been developed.
The U.S. Copyright Office, in its recent guidance on works containing material generated by artificial intelligence, adopts a broader definition of AI-generated material, encompassing any artistic work where AI plays a part in the creative output.
For practical purposes, the distinction between AI-generated and AI-assisted works can be blurry, and we are currently dealing mainly with AI-assisted works, where both humans and AI systems contribute to the creative process. In this blog post, we will use the terms "AI-generated" and "AI-assisted" works interchangeably, as all AI-generated works currently available still involve human input at some level.
The US Perspective: Emphasizing Human Authorship
Historically, intellectual property policies in the U.S. and the EU have evolved slightly differently. While the principles are not fundamentally distinct, the role of registering a copyright is more beneficial in the U.S. Therefore, the rulings of the Copyright Office are essential for understanding the copyright regime in the country.
A recent landmark ruling by the Copyright Office provides guidance on the future of copyrights for AI-assisted works. Partially triggered by the copyright application for Kris Kashtanova's AI-illustrated comic book, Zarya of the Dawn, the Copyright Office explained what can and cannot be copyrighted in the U.S.
The first and most important aspect is that only human authors can claim copyright. The Office's position is that copyright can protect only material resulting from human creativity. The terms "author" in both the Constitution and the Copyright Act exclude non-humans.
In the Office's view, it is well-established that copyright can protect only material that is the product of human creativity. Most fundamentally, the term “author,” which is used in both the Constitution and the Copyright Act, excludes non-humans.
In the context of AI-generated works, the Copyright Office emphasizes the need to determine the role of a human in the creation of the works on a case-by-case basis. If an AI technology determines the expressive elements of its output, the generated material is not a product of human authorship and, thus, cannot be copyrighted.
However, this does not automatically mean that no AI-generated works can be copyrighted. The Office clarifies that, in some cases, a work containing AI-generated material will also contain sufficient human authorship to support a copyright claim. For example, a human may select or arrange AI-generated material in a sufficiently creative way that the resulting work as a whole constitutes an original work of authorship. Although this broadens the range of copyrightable AI-generated material, the Copyright Office also instructs AI creators that using a process of fine-tuning prompts does not qualify their works for copyright. The Office states, "Some technologies allow users to provide iterative “feedback” by providing additional prompts to the machine. For example, the user may instruct the AI to revise the generated text to mention a topic or emphasize a particular point. While such instructions may give a user greater influence over the output, the AI technology is what determines how to implement those additional instructions."
In summary, direct outputs from AI are most likely not copyrightable. However, if an author uses that material and composes something out of it without direct AI assistance, then this could potentially meet the criteria of sufficient human authorship. Additionally, when AI merely plays an assistive role (e.g., correcting grammar), it is more likely that the works are copyrightable.
As we have explored the U.S. approach to copyright for AI-generated works, it is evident that human involvement plays a crucial role in determining whether a work can be copyrighted. With this in mind, let's now examine the European Union's perspective on the matter and see how their approach may differ from or align with that of the U.S.
The EU Outlook: Tracing Creativity in AI-Generated Works
In the European Union, there is no need to register a work to secure copyright protection. As soon as a work is created and fixed in a tangible form (e.g., written down, recorded, etc.), it automatically receives copyright protection. This principle is in line with the Berne Convention for the Protection of Literary and Artistic Works, which most countries, including EU member states, have adopted.
While some EU countries have voluntary copyright registration systems that can serve as evidence of copyright claims, registration is not mandatory for copyright protection. Instead, the focus is on the originality and creativity of the work, which are the main criteria for a work to be eligible for copyright protection under EU law.
In their article "Copyright and Artificial Creation: Does EU Copyright Law Protect AI-Assisted Output?", Hugenholtz and Quintais suggest a 4-step structure to evaluate whether a work can be qualified for copyright in the EU:
- Does the work fall into the literary, scientific, or artistic domain?
- Is there human effort involved?
- Is there originality and creativity (including human creative choice)?
- Does the work "express" the human creator's creativity in the final product?
The authors also highlight the importance of understanding that there is no single EU-level directive that harmonizes the concept of work of authorship. Consequently, the law relies mostly on the Berne Convention's definition of literary and artistic works that can be protected by copyrights.
Generative AIs are capable of producing works that fall into the literary, scientific, or artistic domain and therefore cannot be excluded from copyright protection on that ground alone. So we need to move on and verify the human presence.
Similar to the approach of the Copyright Office, works can be copyrighted if it is possible to prove human authorship. While it is clear that the output of AI that did not require any human input is not copyrightable, the extent of human involvement required is not quite clearly defined in the law. If a human was present in the creation, the work may qualify for copyright protection. The Court of Justice of the European Union (CJEU) has said that it is possible to create works of authorship with the help of a machine.
The third evaluation criterion is originality or creativity (creative choice), and the key question here is whether the author was able to express their creative abilities freely in the production of the work. This was emphasized in the Painer case ruling by CJEU. The Painer case illustrates that creative choices may occur at various levels and in different phases of the creative process: preparation, execution, and finalization.
Generative AI is capable of being involved in every step of the creative process equally. It's more than likely that the human component is more strongly emphasized in the steps of the preparation phase, for example, by defining the goal of the works, general artistic boundaries, styling, etc. While generative AI's role is stronger in the execution phase by producing the results based on the creative direction given. Finally, in the last phase, it's most likely the human who judges whether this work can be called "final" or if it should be iterated further or improved in some other way.
However, as generative AI becomes more powerful and with the support of reinforced learning, it is still possible to imagine that human input in works is minimal or close to non-existent. Also, copying the prompt and regenerating it may lack the expression of originality and creativity needed to claim authorship of the work.
The fourth criterion is whether the author's creativity is expressed in the works. To establish this, we need to determine the original authorial intent and how it is linked to the final work. While generative AI follows statistical probabilities to reach the output work, the output is not completely random and can, in most cases, be linked back to the original intent encapsulated in the prompt and additional data provided as input. Thus, just because an AI system is treated as a 'black box'—with its inner workings not fully understood—this alone should not be grounds for disqualifying a work from copyright protection, as long as human input and creative choice are present in the creation process.
To summarize, the EU approach to AI-generated works focuses on the presence of human authorship, the originality and creativity of the work, and the expression of the human creator's creativity in the final product. Although the legal framework is not explicitly clear on the extent of human involvement required, works involving some level of human input and creative choice may still be eligible for copyright protection under EU law. As the field of AI-generated works continues to evolve, legal and ethical discussions will play a crucial role in shaping the future of copyright protection in the European Union.
Comparing Approaches: US and EU Copyright Laws Confront AI
The US and EU approaches to copyright protection for AI-generated works share some similarities, but there are also important differences. Both jurisdictions emphasize the importance of human authorship and creative input in determining copyright eligibility. The main distinction lies in the extent to which the human involvement is required and how it is assessed.
In the US, the Copyright Office has made it clear that only works of human authorship can be copyrighted. As such, AI-generated material cannot be copyrighted if the AI system solely determines the expressive elements of the output. However, if a human creator selects or arranges AI-generated material in a sufficiently creative way, the resulting work may be eligible for copyright protection.
The EU approach, while not as explicit as the US, also focuses on the presence of human authorship, originality, and creativity in the work. The EU courts have recognized the possibility of creating works of authorship with the help of machines. Still, the extent of human involvement required for copyright eligibility is not clearly defined in the legal framework.
Another difference between the two jurisdictions is that the US places a greater emphasis on the registration of copyrights, while the EU offers automatic copyright protection as soon as a work is created and fixed in a tangible form. This means that, in the EU, the focus is on the originality and creativity of the work rather than the formal registration process.
In conclusion, both the US and EU approaches to copyright protection for AI-generated works highlight the importance of human input and creative choice. As AI technology continues to advance, legal and ethical debates surrounding the ownership, authorship, and originality of AI-generated works will be crucial in shaping the future of intellectual property protection in both jurisdictions.
The Road Ahead: Copyright Challenges in the Age of AI
As we have explored in this blog post, the rapidly advancing AI technology is challenging the traditional principles of intellectual property and copyright protection. While the US Copyright Office has provided some guidance on the treatment of AI-generated works, many questions and uncertainties remain. Similarly, the EU approach to copyright protection for AI-generated works is still evolving, and the application of existing legal principles to AI-generated works is yet to be seen in real-life cases.
Both the US and EU approaches emphasize the importance of human authorship, originality, and creative choice in determining copyright eligibility. However, as AI technology becomes more sophisticated and capable of producing increasingly complex and creative works, it is likely that legal and ethical debates surrounding the ownership, authorship, and originality of AI-generated works will continue to shape the future of intellectual property protection in both jurisdictions.
In the meantime, creators utilizing AI-assisted tools should remain vigilant and informed about the ongoing developments in copyright law. As the landscape of AI-generated works continues to evolve, it is crucial for artists, lawyers, scholars, and policymakers to engage in open dialogue and collaborate to ensure that the intellectual property system remains robust, adaptable, and capable of protecting and promoting human creativity in the age of artificial intelligence.
