AI has taken the world by storm, and a plethora of new legal questions have come along with it. Issues relating to AI have only just begun to make their way to the courts, so it may be a while before we have concrete answers, but that doesn’t mean we can’t consider the implications in the meantime.
AI was the central topic of Ropes & Gray’s 6th Annual International IP Summit at BC Law’s campus last month. While the functions AI can serve are almost too many to discuss, one especially interesting application of AI is to generate art. The song “Heart on My Sleeve” used AI to emulate the voices and styles of Drake and The Weeknd with eerie accuracy, and the photo that took first place the Sony World Photography Awards was later revealed to be the product of AI.
The popularity of AI-generated art is already having an impact on copyright law. I had a chance to speak with Joseph Liu, accredited scholar and copyright professor here at BC Law, to break down some of the problems posed by AI in the copyright space.
Professor Liu identified three main clusters of issues: infringement claims based on the use of copyrighted works to train AI, works created by AI that resemble existing works, and the question of copyright ownership over AI-generated works.
Most of the suits now going before the courts are related to the first issue. In order for an AI to generate images, music, and other kinds of creative works, it first has to learn from those that already exist. AI programs accomplish this by scraping vast volumes of works from the Internet, many of which are copyrighted, so the question becomes whether copying those works is allowed under the doctrine of “fair use” – a legal defense that excuses the unlicensed use of copyrighted works in certain circumstances.
The case Author’s Guild v. Google, Inc. dealt with a similar kind of issue, where Google copied a vast number of books to create its online database called Google Book Search. The Second Circuit Court of Appeals held that Google was entitled to a fair use defense because while the books were certainly being copied, they were being used for a different purpose – to provide search results – and transformative purposes weigh heavily in favor of fair use. A similar rationale could apply to AI, but Professor Liu pointed out that the purpose of copying in these cases is to train the AI to ultimately to create similar works, so in that sense, fair use might not be applicable.
The second issue, relating to when an AI-generated work resembles an existing work, is more of a gray area. If it can be established that the AI learned from a copyrighted work to generate its own creations, there may be a stronger case for infringement from that work’s copyright owner. However, an artist’s general “style” is not something that copyright law traditionally protects. Several AI programs have put up guardrails that prevent users from prompting the program to create a work in a particular artist’s style, but whether style-copying could even create a valid claim remains to be seen.
The final issue is determining who is the proper copyright owner of an AI-generated work – the user who enters the prompt, the company or individual who owns the AI, or no one at all. The Copyright Office historically only recognizes copyright in works authored by a human being. Even though non-human entities like corporations can own copyrights, that is only possible through the “work made for hire” doctrine, which requires a human author to create the work and then assign the rights to the corporation.
Recently, a man named Stephen Thaler sued the U.S. Copyright Office for denying his application to copyright an AI-generated image. The U.S. District Court for the District of Columbia ruled in favor of the Copyright Office, reaffirming the notion that a work requires a human author to be eligible for copyright. While authors are free to use tools like cameras to create copyrightable material, the difference with AI is that the generated image is wholly unpredictable. Even with an incredibly detailed prompt, the user does not know for certain what the AI will conjure up. The user can edit the product, but their copyright options may then be limited to what they personally contributed to the work, such as the prompt and alterations. Professor Liu theorized that a form of joint authorship between the AI’s user and owner could be possible, but no matter who is entitled to the copyright of such works, there are broad policy implications to consider.
It is also possible that AI-generated works will simply have no copyright owners, but such a result may leave many creators unsatisfied. After all, the purpose of copyright law is to incentivize people to create art, so AI could be a very positive development in that it drastically reduces the labor and costs of creativity. However, others may see generative AI as an unfair shortcut, so rewarding those who use it could have the adverse effect of de-incentivizing artists who create works through traditional means.
While much remains uncertain, one thing is clear – it will be fascinating to watch the courts and the legal community at large continue to respond and adapt to the advent of AI.
Eddie Godino is a second-year law student at BC Law. Contact him at godinoe@bc.edu.
BC Law: Impact 