The use of artificial intelligence (AI) in artistic creations has launched major rights questions across the artistic spectrum. Where does AI go to scrape for the knowledge used in responding to AI prompts to create new content? How many works or catalogues does AI need to ingest to respond effectively to a prompt? What constitutes fair use, or substantial similarity and originality, when a work is mixed into numerous pieces of data? As AI continues its rapid development, creative industries and artists must keep pace to address issues of misappropriation and infringement that threaten artistic control and the integrity of their artistic persona and creations.

On June 24, the Recording Industry of America (RIAA) filed a copyright infringement lawsuit on behalf of the major record labels (Universal, Sony, Warner and their subsidiary labels) against defendant music generative startups Suno and Audio.¹ The RIAA claims that the musical data scrapped and ingested for AI generators to create AI musical content derives from the major labels’ copyrighted works. The AI startups counter that such works are all readily available on the internet and can, therefore, be protected through a fair use defense. The RIAA claims that the resulting generative AI composition and recording could not have been created without ingesting the original copyrighted works and that they are meant to replace the original works. Therefore, the fair use defense is not available.

Regardless of the outcome of the pending RIAA claim, independent artists and music companies will still find themselves battling AI to protect and monetize their original voice and intellectual property creations. Artists have historically protected the right to their artistic identity through rights of publicity claims used to protect an individual’s commercial use of their voice, as well as copyright claims used to protect an original specific work. While AI is relatively new, the concept of harm through misappropriation or infringement of an artist’s name, likeness, voice and catalogue is not new at all. Past precedents and developing legislation offer a preview of the defenses and challenges in the post-AI world where protecting against the misappropriation of an artist’s unique sound, including their voice, likeness, music, lyrics or performance, is paramount.

Recent Challenges Arising From AI

Drake v. Tupac

In a recent highly publicized conflict, the rap artist Drake used AI to develop a verse using the vocal stylings and sound-alike voice of late rap artist Tupac Shakur, who passed away in 1996. The song was titled “Taylor Made Freestyle” and was part of a series of diss tracks between Drake and rap rival Kendrick Lamar. The record was released to the streaming services on April 19. Tupac’s estate, which controls Tupac’s rights of publicity and has approval rights on catalogues use, issued a cease-and-desist letter demanding the takedown (removal) of the song. Tupac’s estate had not granted Drake license or any permissions to replicate the voice or access the catalogue. Tupac’s estate theorized that without accessing Tupac’s catalogue and copyrighted works, there was no way for Drake to have used AI to create a Tupac sound-alike verse. The use of Tupac’s voice alone was considered by the estate as a particularly egregious misappropriation; calling it a “flagrant violation of Tupac’s publicity and the estate’s legal rights,” and that the misappropriation was “a blatant abuse of the legacy of one of the greatest hip-hop artists of all time.”² As a result, one week later on April 26, Drake took the song down prior to any further legal action by the estate.

Johansson v. OpenAI

Actor Scarlett Johansson provided the voice recordings as an AI-developed character called “Samantha” for the Spike Jones-directed film Her, which resulted in an identifiable vocal persona. OpenAI introduced “Sky” as one of the new voices for their ChatGPT chatbots, and most people who know Johansson’s work would recognize the voice of “Sky” as Johansson’s voice. After Johansson’s attorneys made several inquiries, the AI-generated voice that sounded very similar to Johansson was dropped from production by OpenAI, although their CEO vehemently denied using Johansson’s voice in “Sky’s” development.³ Although Johansson has not decided whether to file suit (the company halted the use of the voice of the product and said this was only for a not-for-profit product demo), Johansson would have claims arising through rights of publicity and copyright.

Right of Publicity Claims

An original artist may be able to utilize a right of publicity claim to protect their artistic work from misappropriation stemming from the use of AI. Although the elements vary across states, to establish a right of publicity claim, a plaintiff must generally prove the following:

• The defendant’s use of the plaintiff’s identity.
• The appropriation of plaintiff’s name or likeness to the defendant’s advantage
• Lack of consent.
• Resulting injury.

Two major California precedents that are useful when considering rights of publicity claims and AI are Midler v. Ford Motor Co.⁴ and Waits v. Frito-Lay Inc.⁵ These two California cases emphasize that the identity required for a right of publicity claim can also be found within a unique human voice. This line of cases subsequently resulted in California Right of Publicity legislation protecting an artist’s unique voice along with name and likeness.

In Midler v. Ford Motor Co., Ford Motor Company’s advertising agency, Young & Rubicam, Inc., used a “sound-alike” to sing Bette Midler’s “Do You Want to Dance” on a television commercial after they failed to get the nationally known actress and singer to perform the song for a car commercial. The “sound-alike” performer imitated her vocal characteristics and unique style. Midler’s name and likeness were not used in the advertisement, and a copyright license was obtained to use the song. The issue before the court was whether Midler’s voice was protected from appropriation. The court held that Midler’s distinctive voice was protected from appropriation because “the human voice is one of the most palpable ways identity is manifested.” Additionally, the court reasoned that the defendants’ previous request for Midler to sing in the advertisement gave meaning to the value of her voice.

In Waits v. Frito-Lay Inc., the iconic musical artist Tom Waits sued Frito-Lay Inc. after their radio commercial featured a vocal performance that imitated Waits’ idiosyncratic voice and gravelly vocal stylings. The issue before the court was whether the company misappropriated Waits’ voice. The court awarded Waits compensatory and punitive damages because there was a “high probability” of misappropriation. The court referenced Midler’s case to emphasize the identity found within a voice and stated that “what is put forward as protectable here is more personal than any work of authorship . . . a voice is as distinctive and personal as a face.”

The trend to protect an artist’s unique voice has been expanding to other districts and through state legislation. In Prima v. Darden Rests. Inc., the defendants broadcasted a commercial that used the plaintiff’s deceased husband’s vocals.⁶ The issue before the court was whether the defendants infringed the right of publicity through the husband’s identity. The court held that the plaintiff established a prima facie case of infringement because the plaintiff could show that there was an identity within her husband’s voice, the defendants did not have permission to use the vocals and the defendants inflicted damage towards the plaintiff. Citing Midler and Waits, the court stated, “The trend among courts addressing the issue is that imitation of another’s voice can give rise to a cause of action for infringing the right of publicity.” Despite defense claims that these voice claims should be prosecuted through the copyright act, state remedies for rights in publicity have continued as distinct and separate claims.

Another example citing Miller and Waits is in the 5th Circuit as seen within Brown v. Ames, where Collectables, a music recording company, used the artist Roy Ames sound-alike voice to help market and sell his recordings.⁷ The issue before the court was whether the plaintiff musician’s state claims of misappropriation were preempted by the Copyright Act. The court cited Midler and Waits and held that “the Copyright Act did not preempt a claim of misappropriation of the singer’s vocal style” because “vocal style is not ‘fixed’ and therefore it is not ‘copyrightable.’” The court further reasoned that “Midler foreshadows the result in this case, reasoning that ‘a voice is as distinctive and personal as a face.’”

State and Federal Legislation Protecting Rights in Publicity

Many states have enacted rights of publicity laws to protect artistic name and likeness. In Tennessee, the newly passed Ensuring Likeness Voice and Image Security Act (ELVIS Act) specifically includes an artist’s voice as a component of their likeness.⁸ While the law seems appropriate in a fast-paced AI world, critics claim that issues regarding the Tennessee law will emerge because it could be applied in numerous situations it wasn’t intended for, such as a concert attendee’s personal photograph of a performer sent to friends. However, the law that strictly enforces AI misappropriation would also have some potential to protect the public from inadvertently purchasing AI-fabricated fraudulent voice clips, likenesses or music that the artist did not perform.

There are developments on the federal level towards protections for voice and likeness against AI; however, they have been moving slowly due to challenges from big tech and First Amendment defenders. On Jan. 10, a bill called the No Artificial Intelligence Fake Replicas and Unauthorized Duplications Act (No AI FRAUD Act) was introduced with hopes of increasing protections against AI.⁹ Michael Huppe, the CEO of SoundExchange, stated that “the No AI Fraud Act will create a federal property right for likeness and voice while providing real penalties for companies and individuals who attempt to steal and profit off the intellectual property of music creators.”¹⁰ Additionally, a Senate bill called The No Fakes Act was introduced in the fall of 2023 to provide for safety in the use of AI by requiring companies to assess AI technologies before purchasing them.¹¹ Hearings on both bills are ongoing; however, with competing interests, the passage of AI regulations on the federal level appears to be a long way off.

Protection Through the Copyright Act

In addition to a right of publicity claims protecting voice, name and likeness, as stated above, artists with copyrighted or copyrightable original works may also be able to succeed on a copyright claim. To succeed on a copyright claim, the plaintiff must first establish (1) ownership of a valid copyright and (2) copying of original elements of the work. In cases of copyright infringement and AI, the difficulty comes with proving elements of originality and substantial similarity when AI is combining works from multiple data sources. While laws regarding copyright and AI are sure to develop as courts try to keep pace with new advancements, there are a few precedents that predict where copyright protections in this area are headed.

In Broad. Music Inc. v. Joe’s Edelweiss LLC, plaintiff BMI owned the copyrights of works that the defendant’s business establishment publicly performed without permission.¹² The issue before the court was whether BMI had a substantial claim of copyright infringement against the defendants. The court held that BMI did have a copyright claim against the defendants because the plaintiffs could establish:

1. The originality of the compositions involved.
2. Compliance with securing a copyright under Title 17, United States Code.
3. That they were the proprietors of the copyrights.
4. That the compositions were performed publicly by the defendant.
5. That the defendant did not receive permission from the plaintiff.

In Music Publishers v. Anthropic PBC, a developing music industry case within the Middle District of Tennessee, major music publishers claimed that Anthropic PBC, an AI software developer, violated the Copyright Act when it trained a chatbot using at least 500 copyrighted lyrics.¹³ The plaintiffs are claiming that these lyrics will be used to produce “substantially similar” works if the chatbot is asked to write a song. This case is at an early stage; however, precedents and ongoing AI and copyright legislation may help the court to further develop the law within this new area, especially when considering the copyright elements of substantial similarity and originality.

The basic elements of a copyright claim, as set forth in the cases above, become more complicated when adding AI into the equation. How does one establish “substantial similarity” when AI software is gathering parts of songs from multiple sources of data?

In Andersen v. Stability AI Ltd., the plaintiffs alleged that the defendants violated copyright and right of publicity claims when the defendants used the plaintiffs’ artwork to train AI systems.¹⁴ The issue before the court was whether the plaintiffs’ copyright claim could sustain. The court held that the copyright claim could not stand because the AI product and the artist’s work had no “substantial similarity.”

Can an artist still establish originality when their vocals and chords are mixed into this complex data, possibly with thousands of others? To ensure the success of a copyright claim, artists will need to be able to clearly establish originality and substantial similarity within an AI-produced work. With developing technology, we can track and trace the use of copyrighted works in the scraping and ingesting of AI data, which should aid in future copyright claims and offer protections against fake works created by AI.

Fair Use Defense

AI companies have attempted to use a fair use defense to combat the power of a copyright claim against AI. Music Business Worldwide stated, “like with the RIAA-led lawsuits against Sudo and Udio, much will hinge in these lawsuits over whether the courts accept the notion that using copyrighted materials to train AI constitutes ‘fair use’ or is a violation of copyright law.”¹⁵ The fair use defense requires a balancing of factors including (1) the purpose and character of the use, (2) the nature of the copyrighted work, (3) the amount of substantiality of the portion used in relation to the copyright as a whole and (4) the effect of the use upon the potential market for or value of the copyrighted work.

While the courts are still deciding on this issue, precedents such as Authors Guild v. Google, Inc. are likely to play a role in the decision.¹⁶ In Authors Guild v. Google, the plaintiffs sued Google for making digital copies of millions of the plaintiff’s books. The issue before the court was whether the fair use defense applied. The court held that the fair use defense did apply because the copying was transformative, and there was no significant change in market value for the products.

An example of where the fair use defense did not apply is seen within Hachette Book Grp. Inc. v. Internet Archive.¹⁷ Here, the defendants scanned copies of the plaintiff’s books without their permission. The issue before the court was whether the defendants had a fair use defense. The court held that the copying of the plaintiff’s work was not fair use because fair use does not allow for mass reproduction without transformation of the works.

Courts have been wary over just how far the fair use defense can go. In the instance of AI, there may not be as strong of a fair use defense when considering that AI-produced works are derived solely from ingesting the data of the original works and often can produce close versions or substantially similar versions of the original works. A multitude of new and similar works can change the market value of the song and cause financial harm, two major elements of copyright infringement.

The Takeaway

Today, AI is readily available to the public and can potentially misappropriate the voice and catalogue of any artist. This is significantly different from the era of the late 1980s and early 1990s, where a human still had to read the part and act, speak or sing in the distinctive style of another, and such performance had to be professionally recorded. OpenAI can now create this interpretation or appropriation on command. But for AI to create this “performance,” AI must ingest or “scrape” one or multiple artists’ voices and creative works to spit out an exact replica or creation based on the requests in a ChatGPT prompt. If AI is using the copyrighted works and the artist’s distinctive voice, then there is both a potential copyright claim and a rights in publicity claim against a public release or commercial use of the newly created AI performance or track. Artistic creations originate with creative humans, and much of human creativity in literature, music and fine arts is available online. AI is not operating in an artistic vacuum.

The rulings in the line of cases cited above in both state’s rights of publicity claims and federal copyright claims offer plaintiff’s attorneys guidance in addressing the challenges in the complex world of AI as they seek to protect their client’s creative intellectual property. These cases are meaningful to today’s discussion of AI and the complexity involved in pinpointing what constitutes and defines the artist’s unique voice and style and copyright protections for their creative works. While the early cases focused more on commercial advertising, today’s AI world expanded both the use of and the potential for immediate distribution of an artist’s likeness on streaming services and social media.

Ultimately, the key to controlling artistic catalogues and compensating artists for the use of their copyrighted work lies in the tracking of original works either through detective work in determining the origin or through watermarking or other branding technologies. New technology is becoming available to watermark or otherwise imprint original content so that the use, even in micro fractions of a work ingested by AI, can be tracked to the new work. This will enable an artist to demonstrate that their work contributed to a collective work and that the owner of a copyright or license has a financial interest in any new work produced from an AI project, thus requiring permission for the use of the original work, and a negotiated compensation for the holder of the copyright or license. |||

---

HILLEL FRANKEL is of counsel with McGlinchey Stafford PLLC and works out of their Nashville office, focusing on intellectual property and transactional music and entertainment matters. Frankel has over 25 years of experience representing artists, writers, producers, production companies, record labels, visual artists and filmmakers. He also provides counsel in business affairs and business development for start-up production, music tech and recording companies, as well as for apparel, branding, sponsorship and marketing entities. This article was developed with input from McGlinchey’s Summer Associate Kayleigh Thomas.

---

NOTES
1. UMG Recordings Inc., Capitol Records LLC, Sony Music Entertainment, Atlantic Recording Corporation, Atlantic Records Group LLC, Rhino Entertainment LLC, The All Blacks U.S.A. Inc., Warner Music International Services Limited, and Warner Records Inc. v. Suno Inc. and John Does 1-10, www.riaa.com/wp-content/uploads/2024/06/Suno-complaint-file-stamped20.pdf.
2. Matthew Strauss, Pitchfork, “2Pac’s Estate Threatens Drake With Legal Action Over ‘Taylor Made Freestyle,’” https://pitchfork.com/news/2pacs-estate-threatens-drake-with-legal-action-over-taylor-made-freestyle, (April 24, 2024).
3. Todd Spangler, Variety, “Scarlett Johansson Says She Was ‘Shocked’ and ‘Angered’ Over OpenAI’s Use of a Voice That Was ‘Eerily Similar to Mine,’” https://variety.com/2024/digital/news/scarlett-johansson-responds-shocked-angered-openai-chatgpt-her-1236011135, (May 20, 2024).
4. Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988) https://law.justia.com/cases/federal/appellate-courts/F2/849/460/37485/.
5. Waits v. Frito-Lay Inc., 978 F.2d 1093 (9th Cir. 1992) https://casetext.com/case/waits-v-frito-lay-inc.
6. Prima v. Darden Rests. Inc., 78 F. Supp. 2d 337 (D.N.J. 2000) https://law.justia.com/cases/federal/district-courts/FSupp2/78/337/2568975/.
7. Brown v. Ames, 201 F.3d 654 (5th Cir. 2000) https://casetext.com/case/brown-v-ames.
8. Ensuring Likeness Voice and Image Security Act (ELVIS Act), HB2091/SB2096, https://wapp.capitol.tn.gov/apps/BillInfo/Default.aspx?BillNumber=HB2091, (July 1, 2024).
9. No Artificial Intelligence Fake Replicas And Unauthorized Duplications Act of 2024 (No AI FRAUD Act), H.R.6943 — 118th Congress (2023-2024), (Jan. 10, 2024), www.congress.gov/bill/118th-congress/house-bill/6943/text?s=1&r=9.
10. Kristin Robinson, Billboard, “House Lawmakers Unveil No AI FRAUD Act in Push for Federal Protections for Voice, Likeness,” (Jan. 10, 2024), www.billboard.com/business/legal/no-ai-fraud-act-congress-federal-law-explained-1235578930.
11. Nurture Originals, Foster Art, and Keep Entertainment Safe (No Fakes) Act, www.coons.senate.gov/imo/media/doc/no_fakes_act_one_pager.pdf.
12. Broad. Music Inc. v. Joe’s Edelweiss LLC, No. 3:11-cv-919-J-37TEM, 2012 U.S. Dist. LEXIS 128718 (M.D. Fla. May 10, 2012) https://casetext.com/case/broad-music-inc-v-joes-edelweiss.
13. Music Publishers v. Anthropic PBC, https://casetext.com/case/concord-music-grp-v-anthropic-pbc-1.
14. Andersen v. Stability AI Ltd., No. 23-cv-00201-WHO, 2024 U.S. Dist. LEXIS 22837 (N.D. Cal. Feb. 8, 2024), https://casetext.com/case/andersen-v-stability-ai-ltd.
15. Daniel Tencer, Major Record Companies Sue AI Music Generators Suno, Udio for ‘Mass Infringement’ of Copyright, Music Business Worldwide, www.musicbusinessworldwide.com/major-record-companies-sue-ai-music-generators-suno-udio-for-mass-infringement-of-copyright, (June 24, 2024).
16. Authors Guild v. Google Inc., 804 F.3d 202 (2d Cir. 2015) https://casetext.com/case/guild-v-google-inc-1.
17. Hachette Book Grp. Inc. v. Internet Archive, 664 F. Supp. 3d 370 (S.D.N.Y. 2023), https://casetext.com/case/hachette-book-grp-v-internet-archive-5.