Artificial Intelligence and Copyrights – Dilemmas for Both Infringement and Creation
Over the weekend of April 15, 2023, a video comprising a song titled “heart on my sleeve,” featuring the voices of superstar artists Drake and The Weeknd, went viral, reaching several million views. Of course, as was later revealed, Drake and The Weeknd had nothing to do with this work. Instead, the person, going by the handle “ghostwriter,” who published the song revealed that they had written the song but used generative artificial intelligence (AI) to mimic the voices of the two artists. After Universal Music Group, the representative for both Drake and The Weeknd, filed copyright claims, the video was removed from the predominant music streaming services.
This development presents several issues, primarily in copyright but also potentially impacting trademark rights and the right of publicity. First, assuming ghostwriter (and not an AI) did in fact write the song as an original work, it may be protectable as a musical composition.
The performance of the song, on the other hand, raises more intricate questions. One presumes that in order to train the AI to mimic the voices of Drake and The Weeknd, ghostwriter uploaded into the AI system copies of works performed by the two artists. It is this act of copying that would form the basis of a copyright infringement claim, as any work created by the AI system using the artificial voices could be considered a derivative work of the uploaded genuine songs, created without license.
In addition to the copyright issues, other rights may be implicated. Drake and The Weeknd may have claims of unfair competition, in that this mimicking of their well-known voices amounts to a false designation of origin under Section 1125 of the Lanham Act. The California statutory right of publicity, Cal. Civ. Code § 3344, has been held to limit its protection of a person’s voice to their real voice and not an imitation; however, it would not be difficult to foresee a reinterpretation or statutory amendment to overcome that limitation in view of this mischief. In any event, the common law right of publicity available in California and several other states may provide a cause of action.
Copyright infringement is just one side of the coin when it comes to AI; copyrightability is the other. The Copyright Office has recently issued a Registration Guidance clarifying that copyright applicants are under a duty to disclose whether some or all of the elements of a work submitted for registration were created using AI. The Office has stated that only elements created by a human author can be copyrighted. For example, if an application asserts an AI program as the author of a work as a whole, that application will be rejected. This issue is being litigated currently in a suit by computer scientist Stephen Thaler, who is seeking to overcome a Copyright Office rejection of an application for a work entitled “A Recent Entrance to Paradise” that identified Thaler’s Creativity Machine AI system as the author.
In another instance, an individual artist, Kristina Kashtanova, filed for registration claiming authorship of a graphic novel in which Kashtanova had written the text, but the drawings had been generated by an AI program in response to Kashtanova’s prompts. Kashtanova, however, did not disclose this in the copyright application but instead claimed authorship of the entire work. After receiving a registration for the work, Kashtanova revealed the subterfuge on the internet. In response to this, the Office revoked the original registration and instead issued two registrations, one for the text alone, and one for the arrangement of the text with the drawings. The drawings themselves, however, having been generated by AI, were not eligible for registration.
The use of AI to create – or to infringe – copyright interests raises a host of legal questions. Be sure to consult your Vorys attorney with any questions you may have.