Keeping Their AI on the Ball - US Copyright Office Issues Guidance on the Copyright Protection of AI Generated Works

At this point there probably is no one on Earth that has not heard of, or even tested, various AI content generation systems, including ChatGPT, Google Bard and Midjourney. The tools are already impressive and are only slated to get better, so many legislatures and national bodies are (finally!) getting to grips with AI, its implications, and legal issues around the technology. One such aspect is whether AI generated works can, or will be protected by copyright, as the area remains firmly hazy as the law stands, including in the common law. One of the first off the blocks is the US Copyright Office which recently released guidance on the protection of AI generated works. 

The guidance, titled "Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence" is brief and to the point, covering most of the issues to give better light on copyright protection on these emerging works.

The guidance initially starts with the basics, noting that for works to be protected by copyright they must have human authorship. The main decision in this regard is Burrow-Giles Lithographic Co. v Sarony where the US Supreme Court confirmed that  an author is “he to whom anything owes its origin; originator; maker; one who completes a work of science or literature”, which specifically can only be a human author. Subsequent decisions have confirmed the position to be the same, with precedent being clearly set in stone since 1886.

The same requirement is extended to the registration of copyright protected works by the Copyright Office, with its Compendium noting this position clearly. Any works created by AI systems are treated as the result of “mechanical reproduction” instead of an author’s “own original mental conception, to which [the author] gave visible form”, so the authorship of AI generated works will depend on the circumstances, e.g., how the AI tool operates and how it was used to create the final work. 

The Copyright Office notes that this includes simply giving the AI system a prompt and it then produces a new work, which is how most of the systems operate currently, as then the "'traditional elements of authorship' are determined and executed by the technology". What users would have to do is exercise ultimate creative control how such systems interpret any given prompts and then generate the material for them to have human authorship. 

Human authorship is possible, however, if a user "select[s] or arrange[s] AI-generated material in a sufficiently creative way that “the resulting work as a whole constitutes an original work of authorship", or if an artist modifies the works created by the AI systems to meet authorship requirements. 

The guidance is a clear sign that legislation is behind the times and legal requirements for authorship would need to be updated to reflect the new paradigm of AI generation of works and their protection in some form. The protection would still have to balance the remuneration of artists whose works are used in the training data and/or source of any derivative works, while encouraging the development of these AI systems and to remunerate them accordingly. This will be a difficult balance to strike, especially since legislatures across the world are woefully behind the development of these technologies.  

Sine Hominum - The Impact of Artificial Intelligence on Copyright

When people speak about artificial intelligence, or AI (the development of computer simulation of human behavior), they often think of things like Skynet or a near omnipotent, all-knowing computer entity that will overcome human's capabilities and rule us all through their superiority. While sci-fi can be a little heavy-handed on the drama, AI is coming and is already somewhat a part of our lives. As things develop in this field more and more legal question will have to be answered; however, this writer has been pondering its effects on copyright specifically, and what the future might hold, particularly in relation to authorship.

In the UK, under the Copyright, Designs and Patents Act 1988, section 154, for a work to be protected by copyright it needs an author (discussed more here). As the law stands today this would have to be a legal person, and a computer generated piece of work, with no human input bar the basic themes or elements of the work, would have no rights in the work as an author (nor, arguably, its creator). Similar provisions exist in Canada and Australia that also require a human author.

In the United States the situation is not much different, as an author is required under 17 USC 201 (albeit is not defined further), but the US Copyright Office has raised it as a possible future problem as far back as over 50 years ago. Their considerations even back then echo the main question when it comes to copyright and AI: "...whether the "work" is basically one of human authorship, with the computer merely being an assisting instrument, or whether the traditional elements of authorship in the work …were actually conceived and executed not by man but by a machine". The US District Court did look at non-human authorship some time ago in Naruto v Slater (discussed more here and here), and maintained that non-humans cannot have authorship in works, which has been established some time in the US judiciary (specifically referring to 'humans' or 'persons'). It is quite clear therefore that AI wouldn't have authorship under US law, and no copyright would vest in the works created.

AI can make up its own mind

A possible extension of the authorship of the work, to enable to protect the work under copyright, is to view the AI (or computer executing the AI's commands) as an extension of the human author. The UK case of Express Newspapers Plc. v Liverpool Daily Post & Echo Plc discussed the creation of works with computers programs, and determined that "...[t]he computer was no more than the tool by which the...[works] were produced to the instructions, via the computer programmes, of [the author]". This was codified in section 9 of the CPDA 1988, where the provision states that "[i]n the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken". While one could conceive that the AI could be seen as an extension of the human author, it still raises the question of actual input. A very sophisticated AI wouldn't, and at the heart of AI development, shouldn't require human input, bar the very basics of what is meant to be achieved through it. The AI should be able to 'think for itself' and therefore it would be a stretch to call the AI a mere tool of the human operating it initially.

The case of Nova Productions Ltd v Mazooma Games Ltd further adds to this, where the court deemed that the authorship of the copyright protected works would be afforded to the human programmer as they "…devised the appearance of the various elements of the [video] game and the rules and logic by which each frame is generated and he wrote the relevant computer program". One should be cautious with this logic as the creation of an AI that could itself create, with minimal or no input at all, a near limitless amount of protected works, and possibly lead to a virtual monopoly in music, art or whatever the AI creates simply due to its capacity to create works on-demand (or even outside of demand). Clearly, in a broad application of creative AI in this way should be denied protection as the original programmer(s) would simply reap the benefits of their artistic 'creations' and potentially limiting the market based on the AI's capacity to create and do so nearly instantaneously. A distinction therefore needs to be made between computer-created works and works created with the assistance of computers, which would protect the interest of active, computer-assisted creators, but limit rights in passive AI creation in the absence of clear and substantial human input.

The future of copyright and AI seems quite murky, with very little legislative discussion surrounding these issues and their future. Even though changes right now might be premature, the law should be ready to follow the technology as it develops and not only be reactionary as and when needed. In being one of the few to look ahead, the EU Legal Affairs Committee has called for "...a criteria for an ‘own intellectual creation’ for copyrightable works produced by computers or robots", but left the nature of this criteria open. As technologies like Google's DeepDream, Project Magenta and Future University Hakodate's AI that writes novels, the future is almost here, and the law seems potentially unable to adjust or adapt as it stands right now.

AI presents us with a wealth of possibilities, and a way for us to revolutionize how we create, work and interact with our world, but much like the revolution presented by the Internet, AI will drastically change the landscape in which the law has to operate. While it might be possible to extend the current provisions to cover these issues, particularly relating to computer-assisted creation, specific legislation would arguably be needed to prepare, not just for what exists now, but what might exist in the coming decades.