A Certainly Uncertain Future: How U.S. Copyright Law May Treat Works Partially Created by Artificial Intelligence
The Release of ChatGPT
In late November 2022, ChatGPT, a form of artificial intelligence (“AI”) was released to the public.[1] Created by OpenAI, ChatGPT is a search box that allows users to pose almost any sort of question and receive a conversational, cogent response within moments.[2] Users can ask for answers to math equations, pre-written email text with particular information, code for a new sort of video game, a poem about Peru in iambic pentameter… the options are endless.[3] Five days after its release, ChatGPT accumulated approximately one million users.[4] Within two months, the program reached 100 million monthly users, making the record for the fastest-growing application in history.[5]
ChatGPT arguably has the ability to generate entire creative works–art, music, literature– “better and faster than humans.”[6] But what if authors and artists choose to use ChatGPT only as a tool to fill in the gaps of a work that they otherwise created from scratch? Or, what if they use ChatGPT to create the base of a text or a project, and act merely as an editor, adding sentences and changing paragraph structure, but keep the AI-created base the same?
The Basics of U.S. Copyright Law
The Copyright Act of 1976 (“Copyright Act”) governs the protection of original works of authorship, whether literary, musical, or artistic.[7] The purpose of copyright is to encourage creative expression by granting authors exclusive rights to their works for a limited period of time.[8] In general, only the copyright owner has the exclusive rights to reproduce, distribute, perform, display, and prepare derivative works based on the original work.[9] However, there are certain situations where someone other than the author may have the right to claim the copyright in a work.[10] For example, if a work is created by an independent contractor and the contract specifies that the copyright in the work will be transferred to the commissioning party, the commissioning party will hold the copyright.[11]
The originality requirement, set forth in Section 102(a) of the Copyright Act, is a central piece of the copyright law regime.[12] It states that copyright protection is available for “original works of authorship fixed in any tangible medium of expression.”[13] This requirement means that in order to be eligible for copyright protection, a work must be original and created by the author.[14] Plagiarism involves copying another’s work without attributing credit to the original source, which typically lacks originality.[15] Thus, a work that is fully plagiarized does not meet the originality requirement for copyright protection (not to mention how it may also constitute copyright infringement).[16] Plagiarism could conceivably include using a text created by AI without proper attribution or presenting it as original work. However, plagiarism in this context is not yet clearly defined.[17]
How Copyright Might Apply to Works Partially Created by AI
A work created wholly by ChatGPT is technically an original work, but is without a human author.[18] Well, sort of. ChatGPT is able to produce answers by drawing on a large amount of data that it was programmed to scrape from all over the internet.[19] Programmers trained ChatGPT with Reinforcement Learning through Human Feedback (“RLHF”), a process by which it was taught to produce extremely human-like and accurate responses.[20] Therefore, to whom does the copyright of a ChatGPT-created text belong? The programmers of the application? They didn’t create the work itself–they only created the program.[21] The owners of ChatGPT? They aren’t the close-up “creators” of the text that the Copyright Act seems to contemplate. The user who inputs the question or prompt into the application? They didn’t write the text themselves, thus failing the originality requirement.
As of now, a user who inputs questions to ChatGPT and assembles a text created completely by the application cannot receive a copyright for the work.[22] However, in cases where the author made a “substantial” contribution to the text, lawmakers have indicated that a copyright may be possible.[23] In September 2022, the U.S. Copyright Office (“USCO”) granted a copyright registration for a comic book created partially by AI, but substantially by its human author, Kristina Kashtanova.[24] Kristina claimed that she “provid[ed] details of [her] process to show that there was substantial human involvement in the process of creation of this graphic novel.”[25] The “substantial human involvement” standard could conceivably be met by an author carefully constructing and reconstructing prompts to feed to the AI and manual input, “suggesting a relatively high degree of intellectual [and therefore original] involvement.”[26]
Instead, lawmakers could choose to grant a copyright to an author for only the portions of a text that were created without the help of AI. However, that would require an author to keep track of each sentence as they formulate a text and would require USCO to parse through a work line-by-line with advanced tools, hoping to get it right. One could only imagine how this process might apply to artwork; the line between AI-generated art and human-generated art in a finished product might be very blurry.
Lastly, lawmakers could choose a simple path: No matter how much of a work is generated by AI, either the author could own the copyright to the entire work outright or the AI application could own the copyright to the entire work outright. However, this would leave out the copyright owners of the works that the AI drew on to create the text or art piece.
No matter how lawmakers decide to approach this, one thing is clear: The process of creating a new copyright law regime to account for AI will likely be anything but easy.
Footnotes
