Whither AI and copyright? Two significant battles working their way through the United States court system will define the relationship between human creators and AI systems. If you are a creator, you need to pay attention.
First, can content created by AI be copyrighted?
While it seems evident that content created by a user prompt should not be copyrightable by the user, what about the designer and operator of the AI system? It might seem reasonable to infer the humans who create a system that, in turn, creates new “works” should be able to copyright those works.
In August of 2023, however, a United States District Court Judge ruled that AI-generated content cannot be copyrighted because “human authorship is a bedrock requirement of copyright.” The human creator of the AI system is not directly responsible for the content the system creates, and since machines do not have rights — including copyright — copyright claims are precluded.
Second, the flip side of this question is, can AI content generators use copyrighted source material, copying the style and substance of a human creator’s work?
So far, in Andersen et al. v. Stability AI Ltd. et al, the answer is yes. While the entire case has not been dismissed, most of it has been. The reasoning is the content creators suing the AI operator have not shown:
- The AI system produces works identical to the creator’s works
- The AI system does not “store” the creator’s works in a way that violates copyright
A creator’s works can be copyrighted, it seems, but not a creator’s style. Whether this division between substance and style stands after a few rounds of court cases remains to be seen.
One place where copyright claims appear solid is over a person’s (or a building’s, such as in the case of the Sagrada Familia cathedral) likeness, including the sound of their voice. No one seems to have sued based on their style of writing or painting, for instance, being a part of their likeness. This line of argument might stand in cases where the primary interaction between a creator and the public is published works rather than their presence. It is difficult to see how the “sound of someone’s voice” cannot be construed as a “style” that cannot be copyrighted while a person’s writing style cannot.
For the moment, from a content creator’s perspective, AI is a sort of copyright no-man’s land. The world of AI appears to be shaping up much like the world of social media. Individual users contribute content, and social media services filter and process this information, selling it back to users for a huge profit.
Just as social media services count on everyone producing at least one piece of quality content now and again that they can resell, AI services may end up just counting on taking enough original material from enough people to continue generating “interesting” content often enough to have a sellable product (all in the name of progress, of course!).
At some point, creators will either get tired of this “deal” and stop creating, or these systems will feed into themselves to the point of creating mediocre nonsense.
One path forward for creators is to find some way to create “un-scrapable platforms” so they can publish work that must be purchased. Whether such a system can be built and maintained is another question — but it is an area ripe for development.