Artificial Intelligence, Creativity, and Copyright

Late last month, the United States Copyright Office (USCO) released the second part of its report on copyright and artificial intelligence (AI). Among other developments, the report detailed the Office’s conclusions on whether AI required a new legal structure for any copyright concerns its usage may produce. As some advocated for and testified to, the USCO has now affirmed that copyright protection in the United States requires human authorship and concluded that current copyright law is a sufficient structure to support AI-generated content and ensure protections for creators.

The USCO makes a number of conclusions in the report but four key conclusions are:

1. The use of AI tools to assist rather than stand in for human creativity does not affect the availability of copyright protection for the output;
2. Copyright protects the original expression in a work created by a human author, even if the work also includes AI-generated material;
3. Whether human contributions to AI-generated outputs are sufficient to constitute authorship must be analyzed on a case-by-case basis; and
4. Human authors are entitled to copyright in their works of authorship that are perceptible in AI-generated outputs, as well as the creative selection, coordination, or arrangement of material in the outputs, or creative modifications of the outputs.

But what do these conclusions mean? The USCO acknowledges that “no court has recognized copyright in material created by non-humans,” for instance a monkey taking a selfie cannot receive copyright. This is no different with AI, humans alone remain the only beings that can receive copyright protection. The report also concludes that humans can generate works with the aid of AI that are eligible for some degree of copyright protection. The question then becomes how do we determine the level or nature of human input that is necessary for a work created with the aid of AI to be copyrightable? And as new use cases develop, further integrating AI into art and other forms of content, what about works created through AI tools using methods we cannot conceive of? To support these conclusions the report draws on multiple court rulings, three of which are outlined below.

First, the report recalls the Supreme Court’s analysis from Feist Publications, Inc. v. Rural Telephone Service Co. here, noting that the Court “made clear that originality is required, not just time and effort.” This is to say that a work created with the aid of AI can be copyrightable if a “creative spark” can be shown. Second, the report references the Court’s considerations in Burrow-Giles Lithographic Co. v. Sarony, a case revolving around whether a work created in part through a machine was copyrightable at all. The report summarizes the Court’s position clearly, “the use of a machine as a tool does not negate copyright protection, but the resulting work is copyrightable only if it contains sufficient human-authored expressive elements.” And third, the report draws on a final Supreme Court opinion in Community for Creative Non-Violence v. Reid, which centered on the necessary degree of human contribution to qualify as authorship. The report notes the Court’s conclusion declaring that “the artist’s contributions, which included sketching the design and executing his creative vision in a tangible medium of expression, made him an author.”

In short, the report states that yes, works created with the aid of AI can be copyrightable but there must be a clear degree of human contribution to an AI-generated output and the work in question must show a human’s creative expression. What a human’s creative expression is in the circumstance of a work created with the use of AI and the copyrightability of that work are both determined on a case-by-case basis. There is a unique idea/expression dichotomy at play in each of these works that requires individual analysis and while AI-enabled works can be compared to each other to determine levels of human control, they cannot be judged in a blanket manner. 

Nevertheless, all of this is not to say that every human-created work that utilizes AI is easily copyrightable. The USCO does make clear that “prompts alone do not provide sufficient human control to make users of an AI system the authors of the output.” Furthermore, AI systems are built on extremely complex models that can be quite difficult to evaluate when attempting to understand the process of input to production. AI often goes beyond a prompt when creating an image, or even doesn’t address part of it. And the same prompt run twice could, and almost always does, give a different result. This is a clear lack of creative control that tips the scale against consideration of copyright. The report outlines an argument from AI proponents, noting that AI works are not dissimilar to a Jackson Pollock painting or a nature photograph in some regards. But as the report notes, both of these examples still require a heavy degree of human execution and determination. Pollock may not choose where the paint lands but he does choose the canvas, paints, layers, colors, composition, and more.

The report concludes with the USCO stating it sees no need for any legal change. AI is still in its infancy and existing copyright principles are sufficient to determine whether AI-generated work reflects sufficient human participation and therefore warrants copyright protection. If in the future, technological advances provide creators with more precise control over AI, then a new legal structure may be needed. However, we can rest assured that works generated through artificial intelligence alone won’t be supplanting human creativity and receiving copyright anytime soon.