Artificial intelligence has begun to change the world, and we are only touching the tip of the iceberg in terms of its potential applications. Increasingly, AI programs are being used to generate creative works. Google’s Deep Dream Generator takes photos submitted by a user and creates a new image combining elements of both images. Different Google artificial neural networks have created works of art generated from white noise, and a piece of music that could top the pop charts if it just had Taylor Swift singing over it. The question that arises is how exactly does copyright law relate to works that are created with limited control by humans or with no human involvement at all.
Going back to the 1980s, Courts have examined the question of who exactly owns the copyright in works that are created with the help of a software program. In cases related to video games with a limited number of outcomes (e.g. Mario defeating Bowser) a game player can never claim copyright in the game. The game creator always owns the copyright in the software code and the audiovisual display associated with it.
In instances in which a user creates something with a computer program intended to assist with creativity (e.g. I draw something with Microsoft Paint), the user has the right to the copyright in the creation provided that the user has performed the “lion’s share” of the creation. If the software program has done the “lion’s share” then it is possible that the creator of the software has a claim to the copyright.
The next step in evolution that we are starting to encounter is when an AI program creates something effectively from scratch. A good example is the aforementioned Google neural network that takes random noise and creates a piece of art from it (which are surprisingly aesthetically pleasing). The short answer is that piece of art is not entitled to any copyright protection and goes immediately into the public domain upon creation. Why, you ask? Because in order to obtain copyright protection a work requires a human author. Copyright law simply cannot compute protecting such works, pun intended. The current Copyright Office Compendium, which sets out the rules and practices of the US Copyright Office, explicitly states ““produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.”
There are very few cases on this subject, but the most prominent one actually required the Court to take a step back in the evolutionary ladder, rather than forward. A photographer left a camera outdoors in a wildlife preserve and rigged it so that it would be easy to hit the shutter. A macaque monkey named Naruto took multiple selfies with the camera, including a fantastic photo in which she appeared to be smiling. The photographer licensed the photo and received compensation for it. People for the Ethical Treatment of Animals promptly sued the photographer, arguing that the monkey, Naruto, owned the copyright in the photograph. The federal district court found otherwise, and held that the monkey lacked standing under the Copyright Act because it wasn’t human. The case was appealed to the 9th Circuit Court of Appeals, but the parties ultimately reached a settlement in which 25% of the future proceeds from licensing the photo will go to charities that protect the welfare or habitat of the macaque monkeys.
The fact that solely AI-created works fall into the public domain is problematic for a number of reasons. First, what is the incentive to create an AI program if the creator can’t benefit financially from its creations, and any third party could use the output for their own purposes? Second, what happens if an AI-created work infringes the copyright of another? Who is the owner of the AI-created work—the creator of the program or the party that licenses the AI to create specific works for it to use? At this point, business and the courts haven’t even started to approach these questions. They will have to, and a lot sooner than we all thought.