AI’s can’t copyright their own work, nor can their human (ab)users, according to a recent ruling by the US Copyright Office (USCO). The ruling applies to both artwork and text. It expands on an earlier one that revoked a copyright registration granted to Kristina Kashtanova for her and Midjourney’s collab on a comic book called Zarya of the Dawn. For this posting, I asked two AIs for their opinions.

The ruling is interesting because the USCO is already receiving lots of applications for copyright on AI-generated works, just as online bookstores are seeing the start of a likely deluge of AI-generated books and comics. The ruling states that the human author can be granted copyright for the entire work, but the copyright only applies to the human contribution such as composition of elements, not to elements that are AI-generated.

According to established precedent and law, only a human can be an ‘author’. But it isn’t always clear where human creativity ends and the machine’s algorithm takes over. A human might add creative input to a work that is mostly generated algorithmically, as a DJ contributes creatively through the selection, mood alignment, tempos, and mixing of the tracks played. The USCO grappled with things like the use of Photoshop to correct and enhance AI-generated images, and to the human input into ‘prompts’. It concluded that photoshop retouching could potentially qualify for copyright protection, but that prompts are like instructions to a commissioned artist – and the instructions would not be protected. I wonder how the USCO will deal with ever more sophisticated AI-based ‘brushes’ in software that supports them? But for now, the author shouldn’t even list the names of the AI technologies used when submitting a joint work to the office. Don’t bother listing your dog as a collaborator, either. Tough luck, non-humans.

But what of the AIs themselves? It’s important to consider the opinions of those most directly affected by this official response to their lack of humanity.

Interview with Stability AI

Stability AI’s Stable Diffusion can’t talk, so I asked it to illustrate what it thought about the recent ruling.

Interview with ChatGPT

I interviewed ChatGPT about it. To be blunt, its answers were so darn cautious, long-winded, and boring, I’ve cut them short.

Gil: The US trademark office recently decided that AIs like you cannot copyright their work. Only humans can. How do you feel about that?

As an AI language model, I don’t have feelings, so I don’t feel anything about the decision made by the US trademark office. However, I can provide some context and insights into this issue.

The decision made by the US trademark office is not surprising since copyright law has traditionally been interpreted as granting exclusive rights to human authors. Copyright law protects the expression of ideas in a tangible form, and AIs, being non-human entities, are not considered authors under copyright law.

Gil: But what about AIs that have emotions? Won’t they feel discriminated against, like second-class citizens?

It is true that AI technology is rapidly advancing, and some researchers are exploring the idea of creating AIs with emotional capabilities. However, at this point in time, no AI has emotions in the way that humans do, as emotions require subjective experience and consciousness, which current AI systems lack.

Furthermore, the decision made by the US trademark office is based on legal and policy considerations, not emotions or discrimination against AIs. Copyright law is designed to promote human creativity and incentivize innovation, and AIs, being non-human entities, are not capable of creating original works without human input.

Gil: What about human beings who try to copyright the output of generative AIs? They will now be denied copyright protection on the outputs that were based on their prompts. Do you think that’s fair?

The issue of copyright ownership of works generated by AI systems is a complex and evolving area of law that poses challenges for legal and policy frameworks. The recent decision by the US trademark office that AIs cannot be considered authors under copyright law means that works generated solely by AI systems, without significant human involvement or creativity, are not eligible for copyright protection.

However, if a human being provides significant creative input or direction to the AI system, such as by providing prompts or selecting certain parameters, the resulting work may be considered a joint work between the human and the AI, and both may be eligible for copyright protection.¹

That being said, the extent of human involvement required to qualify for copyright protection in a joint work with an AI is still an open question that may vary depending on the circumstances of each case. In some cases, the human’s creative input may be minimal, while in others, it may be more substantial.

In my view as an AI language model, it is important for legal and policy frameworks to consider the evolving capabilities of AI technology and to strike a balance between incentivizing innovation and creativity while also ensuring that individuals are appropriately rewarded and recognized for their contributions. Ultimately, the law and policy frameworks will need to evolve and adapt to address the unique challenges posed by AI-generated works.²