Stolen Voices: AI Audio Mimicry and Singers’ IP Rights
By Megan Chambers
Artificial Intelligence (AI) is one of the most hot-button topics in technology right now. For many, the end goal of developing advanced technology such as AI is to free humans from “grunt work” and allow for more time and resources to be spent on relaxation, entertainment, and creative endeavors.[1] For others, however, seeing how closely AI can compete with human creativity is the most fascinating use, incentivizing the use of AI in as many fields as possible.[2]This rapid introduction of AI into artistic fields, however, has brought with it a myriad of new legal questions. For instance, could a singer prevent others from using AI to replicate the sound of her voice to create new songs?
I stumbled across a video on Twitter where AI has been used to create covers of “Envolver” by Anitta and “Kill Bill” by SZA in Ariana Grande’s voice.[3] Do the clips sound perfectly like Ariana Grande? No, I wouldn’t say so, as there are definitely moments where the audio sounds warped. But even so, I must admit that they’re realistic enough that if I hadn’t already known they were AI-generated, I likely would have believed that they were real. The use of AI audio to replicate human speech is relatively commonplace now, but using it to replicate a specific artist’s voice, and to do so successfully, is a wholly new frontier.[4]
Let’s say someone released one of these fake Ariana Grande covers on YouTube, claiming it was a real performance. Or suppose a musician who loves writing music but hates her singing voice used an AI imitation of Ariana Grande to record the vocals, then released it on Spotify as her own original song. Although intellectual property in music is largely governed by copyright, current U.S. copyright law does not grant musical artists the right to have their works properly attributed to them, or to prevent misattribution of works they did not create.[5] This right is only afforded to authors of works of visual art existing in either a single copy or 200 or fewer numbered and signed copies.[6] A famous painter, for example, can thus use §106A of the Copyright Act to prevent someone else from falsely attributing a painting to him, whereas a famous author is given no such right by this or any other section of the Copyright Act to prevent another writer from selling a book she did not write under her name.[7]
Under the present laws, Ariana Grande would have no remedy through copyright to force someone to properly attribute the edited “Kill Bill” audio to AI rather than to her.[8] However, she could likely achieve a remedy through state tort claims, such as fraud or slander. Trying to force someone to stop using an AI imitation of her voice for their own music would be an even more complicated task. Ariana Grande only has copyright in her existing songs, not in the sound of her voice itself.[9] And although one can register certain sound trademarks, this protection is only for specific, usually short sounds, like the NBC tri-tone[10], that are used to help consumers recognize a business.[11] The sound of an artist’s overall “voice,” which itself is hard to define or qualify, is not protected by trademark. Ariana Grande’s best option would likely be to assert a right of publicity claim to control the use of her voice and prevent fraudulent imitations[12]; however, the recognition of this right varies state-by-state.[13]
How these mounting legal questions are resolved will largely depend on if and how Congress chooses to respond to these developments. A federal right of publicity statute or an expansion of VARA would allow greater opportunities for singers nationwide to assert that they have full control over the use and attribution of their voices, including close intentional imitations. In the meantime, artists may be in danger of a tsunami of vocal copycats.
Megan Chambers is from Virginia Beach, Virginia and is a second-year student at Vanderbilt University Law School.
You can download a copy of Megan’s post here.
[1] Q.ai, The Pros and Cons of Artificial Intelligence, Forbes (Dec. 1, 2022), https://www.forbes.com/sites/qai/2022/12/01/the-pros-and-cons-of-artificial-intelligence/?sh=6b4029ba4703.
[2] Ashish Sukhadeve, Artificial Intelligence For Good: How AI Is Helping Humanity, Forbes (Feb. 9, 2021), https://www.forbes.com/sites/forbesbusinesscouncil/2021/02/09/artificial-intelligence-for-good-how-ai-is-helping-humanity/?sh=30f9dc63366b.
[3] @PopCrave, Twitter (Jan. 21, 2023, 1:59PM), https://twitter.com/PopCrave/status/1616888339587805184.
[4] Dov Greenbaum, Microsoft’s new voice mimicking AI VALL-E presents both opportunities and risks, CTech (Jan. 22, 2023), https://www.calcalistech.com/ctechnews/article/hkw9o2kjj.
[5] 17 U.S.C. §106A.
[6] Id.
[7] See id.
[8] See id.
[9] See, e.g., Midler v. Ford Motor Co., 849 F.2d 460, 462 (9th Cir. 1988) (“A voice is not copyrightable. The sounds are not “fixed.”).
[10] The mark comprises the musical notes G, E, C played on chimes, Registration No. 0916522
[11] See Trademark sound mark examples, US Pat. & Trademark Off., https://www.uspto.gov/trademarks/soundmarks/trademark-sound-mark-examples.
[12] See, e.g., Midler, 849 F.2d at 463 (holding that Ford’s use of a Bette Midler impersonator for a series of commercials violated Midler’s rights for appropriation of identity).
[13] Mark Roesler & Garrett Hutchinson, What’s in a Name, Likeness, and Image? The Case for a Federal Right of Publicity Law, American Bar Association (Sep. 16, 2020), https://www.americanbar.org/groups/intellectual_property_law/publications/landslide/2020-21/september-october/what-s-in-a-name-likeness-image-case-for-federal-right-of-publicity-law/.