How Will Copyright Handle AI Art When it Still Hasn’t Figured Out Sampling in Music?
Like many people, I’ve been messing around with AI tools to make images, writing, and music. I’m primarily interested in breaking things. No, I don’t want AI to shatter humanity, as is often feared. Rather, I’m exploring the capacity of AI as a collaborative partner to venture beyond conventional boundaries; to help conjure new ideas.
As I discuss my experiments with people, a common question arises with regard to copyright: “AI art tools scrape images from the internet to learn and replicate artists’ techniques and styles, don’t they? How is that fair?” The answer is yes, these tools learn from past art. It’s fair in the same way that the longstanding tradition of artists ingesting art history and metabolizing those influences into their own work is fair.
So far though, the Copyright Office itself appears to be more concerned about whether a work created with the help of an AI tool can even be attributed to a human creator at all. While AI art tools do enable people to manifest ideas more quickly than in the past, with a machine shouldering a significant portion of the process, there’s not much distinction here either, from past technological advancements in which machines have assumed tasks previously performed by humans.
Consider the emergence of photography in the 1800s: with the press of a button, a machine could now capture a landscape or portrait that could previously have only been crafted by hand. Painters were upset by this development, but we now understand photography as its own art form; composition, context, production technique, etc. The same can be said of AI art.
Why then should AI tools for artistic creation be treated any differently from photography under copyright law? The process of a text prompt becoming an AI-generated artwork is akin to a landscape being transformed into a photograph. If anything, the argument could be made that an artist nuancing prompts to generate images may have even more detailed control over outcomes than a photographer has in their toolkit of aperture, exposure, focus, etc. Either way, the effort exerted to make the work should not be a determinant of what qualifies as copyrightable. Yet thus far, the Copyright Office has rejected copyright claims in every AI art case brought before it.
So I was interested to discover Lawrence Lessig’s suggestion that “Congress could require AI technologies to submit all works into digital registries tied to data establishing provenance and ownership.” By setting standards for such a digital copyright registry, the government could facilitate a system that supports artists, making it easier to identify the rightful owners of copyrighted work and simplify the process of obtaining rights for reuse.
Such a copyright registry could also help resolve other longstanding issues, like ongoing tensions over music sampling. After at least three decades of dysfunction, we still find ourselves in an absurd era where copyright law can have the effect of holding back cultural gems (like De La Soul’s masterworks) from the public while denying artists their due respect and compensation. Outmoded copyright laws and legal battles going back to the early 1990s have effectively stifled musical collage and disproportionately affected artists from marginalized backgrounds — Black creatives fighting for their rights against White copyright holders whose works were derivative of Black artists to begin with. The currently existing laws and precedents were set without an understanding of how digital technology would fundamentally transform content production, reproduction, and distribution. Instead of nurturing creativity, they have become oppressive.
The fact that we’ve not yet rectified copyright laws to better serve artists and the public after all this time leaves me pessimistic. However, the emergence of generative AI seems to be compelling Congress to take action—with a semblance of bipartisanship even—albeit to a limited extent so far. Perhaps there’s cause for optimism that this could be the much-needed impetus to get things back on track, especially if we heed the ideas of scholars like Lessig, who have dedicated nearly 30 years to finding solutions for problems with copyright law.
In the meantime, I’ll keep messing around, trying to break things, and hopefully continue to be astounded by the marvels that emerge.
