The Copyright Dilemma: Implications for AI-Enhanced Product Development

A recent legal decision by United States District Court Judge Beryl A. Howell has sent ripples through the realms of both technology and creativity. The ruling, which states that AI-generated artworks are ineligible for copyright protection, has raised intriguing questions about its potential impact on product management and engineering in the domain of artificial intelligence (AI) and creative technology.

On the one hand, Judge Howell's ruling does offer a degree of clarity. It effectively demarcates the boundary between human-created and AI-generated content, which can be valuable for engineers and product managers. It aids them in assessing the legal risks associated with AI-generated products. But, it also presents a paradox; while clarity can be beneficial, it might inadvertently deter the use of GenAI in creating products. Products and code created with AI may not be copyrightable. And if they are partially generated by AI chunks of them may not be copyrightable. Imagine code created partially with Copilot. Would those parts, if unaltered by humans, be automatically open source? The decision would seem so. Developers are incentivized to use Copilot, and similar code-generation tools, to increase their efficiency, but the laws would seem to disincent the same thing.

Similarly for graphic designers, and others that might create interfaces. Are the parts of those designs created with GenAI not protectable with copyright laws? One might argue that if the whole is partially human-created and partially created via GenAI, the whole would be protected. But as music copyright suits have shown us, parts of works may be separable from others parts. In Bright Tunes Music v. Harrisongs Music, the owners of the Chiffons “He’s So Fine'' successfully sued George Harrison for copyright infringement for a chord progression in “My Sweet Lord.” Note, the alleged infringement is part of the song, not the entire thing, and yet Harrison lost the case and was forced to pay nearly $1.6M in 1976 dollars. (Listen to outline similarities here that are not apparent to the casual listener but are close enough to deem parts of songs separable for copyright purposes.) Hip-hop MCs and rappers who more directly sampled other works might lead to the same conclusion, that their mixing of a part of an old work with a new work, did not create an entirely new work, that therefore that the infringement was separable, just as mixing of GenAI-created non-copyrightable material might be with human-generated copyrightable material. (For the record - pun partially intended - I am not a lawyer.)

The ruling underscores the significance of human authorship, providing guidance for product managers and engineers in understanding how to position their AI-generated products within the confines of copyright law. However, for product developers dealing with AI tools that involve varying degrees of human-AI collaboration, determining the threshold of human input required for copyright eligibility may prove challenging, introducing legal uncertainties.

There is another wrinkle that becomes important. It is not just decisions made on what mixing of GenAI-created content with human-created content mean in terms of copyright will have an important effect on not only particular products and code that is important. What is the value of companies using GenAI to build their products? The same ruling seems to imply that another company can legally copy the portion of the products that were not human-created. If you had some interesting logo, or interface component that you used AI to create, I too could use that in my product. Similarly, I might be able to use some of your source code. Obviously, in both of these cases, I would need access to the creation and be able to identify what was human created and what was not. That is, a company could lose the protection, at least partially, of the protection copyright law affords, perhaps reducing their enterprise value.