By Rebecca Conway, L‘25
The question of artificial intelligence (AI) for image creation and copyright has been lurking with limited case law since GANs (Generative Adversarial Networks) took to the scene in 2014. GANs revolutionized AI by creating a two part system that “learned” a data set of previous art and then checked its own generated image against the dataset to look for differences. The first AI-generated portrait sold in auction fetched over 400k in 2018. The rapid improvement of such technologies and the rise in popularity of DALL-E over the last two years bring to the forefront the questions of who owns the works created by these programs and what happens if one of the programs infringes on another's work.
AI tools may not themselves own or copyright the works they create. While the US’s copyright law does not explicitly mandate a human inventor, courts have found this to be a requirement. In Naruto v Slater the 9th circuit found that a monkey which had taken a selfie could not sue for infringement on the grounds that if the legislature had wanted nonhumans to be able to own copyrights and sue for infringement they should have expressly written it into the statutes.
While it has been largely settled that AI may not own the copyright of its work, it is not clear who would own the copyright. The popular assumption in these discussions is to treat AI as a consumer product, similar to Photoshop or Illustrator rather than as an agent of the coders that built it. For further discussion of the implication of copyright and AI as an agent of a principal that originally created the program see here.
Presuming that AI is a consumer product, the question of whether humans may claim authorship of the works they prompt AI to make depends on the level of input and control a prompter has. Kris Kashtanova’s comic book with AI-generated artwork has been a test case for these arguments and is currently being considered by the US Copyright Office. Kashtanova points out that it was arduous to find the right wording for the image generation and that they came in with a vision and merely used AI as a tool to create their vision.
Additionally, AI programs require the use of past works in order to "learn" how to produce new works. For a work to be copyrightable, the Supreme Court has found it must contain at least a “modicum” of creativity. Suits questioning whether an AI work may be considered creative, as well as whether it is fair use to teach an AI with another's art are currently also being filed.