In a summer filled with a certain sold-out stadium tour and sky-high ticket sales, music and its stars are as popular (and lucrative) as ever. And with the advent of generative AI, individuals without coding experience can now effortlessly produce compelling “deepfake audio” that sounds eerily like music’s most popular artists. This is because generative AI tools have been trained on recording artists’ vast catalogs. Several websites have already popped up offering fans the ability to create songs that sound disconcertingly like the music industry’s biggest stars. The legal road ahead for generative AI and its potential repercussions appears treacherous. The question is…are we ready for it?
Generative AI’s musical capabilities—and the market’s demand for its creations—were on display shortly after the technology’s ubiquitous rise. In April 2023, an anonymous user masterminded the uploading of an AI-generated mash-up featuring megastars Drake and Abel Makkonen Tesfaye (formerly known as The Weeknd); this track became one of the most streamed singles in the country, tricking many into thinking it was authentic, and garnering tens of millions of plays. Despite there being an open issue as to whether there was a copyright violation—it appeared to be an entirely original composition and sound recording, and the author disclosed that the artists were not involved in the making of the song—the artists’ label saw red and demanded the song be removed from streaming platforms, ominously asking the companies on “which side of history” they wanted to be. To avoid any bad blood in this commercially symbiotic relationship, the streaming services complied...this time.[1]
Similar concerns extend to nearly every art. For instance, Greg Rutkowski, a graphic artist known for realistic mythical drawings (think dragons fighting gigantic eagles), has been so frequently imitated on Stable Diffusion (a generative AI tool that creates graphic art), that users ask for works “in the style of Greg Rutkowski” more than for Picasso, by a factor of 20. Mr. Rutkowski is, understandably, concerned that this technology will diminish his art, and bottom line.
In an era where AI can generate works that echo—and with near-precision copy—established artistic styles and characteristics, questions arise about people’s right to control the commercial use of their identity, known as the “right of publicity”. While generative AI’s ability to mimic art, people and events raises a labyrinth of legal implications we know all too well, this latest article will explore the right to publicity and how it intertwines with generative AI.
(a) Right of publicity, generally
The right of publicity allows parties to sue for an “appropriation of the commercial value of [a person’s] identity.”[2] This is a state-law intellectual property right,[3] the infringement of which can be redressed by the commercial tort of unfair competition. The right of publicity can be viewed as the doctrinal cousin to the right to privacy; while the right to privacy protects one’s right to individual privacy that starts at birth (e.g., redressing mental anguish caused by invading one’s privacy)[4], the right of publicity is an earned right that redresses potential damage caused to the commercial value of one’s identity.
A person’s “identity” in this context is commonly referred to as their “persona,” which means the public character of an individual that has commercial value, including name, voice, image, likeness, reputation and other recognizable attributes. Thus, the right to publicity grants individuals control over the commercial exploitation of their “persona,” generally preventing others from using it in a commercial context without their consent.
(b) Applying the right of publicity to generative AI tools
With the rise of generative AI, replication and borrowing of identifiable characteristics will certainly increase. However, courts have developed well-established jurisprudence to navigate these complex issues, which historically arose mostly in cases where a defendant utilized a plaintiff’s likeness in a commercial advertisement. Central to such cases is the determination of whether the advertisement sufficiently captures the plaintiff’s distinctive persona.[5]
For example, Vanna White successfully sued Samsung Electronics after the company ran an advertisement featuring a robot intentionally designed to look like Ms. White. The court noted that while viewed separately the “individual aspects” of the advertisement are insignificant, when viewed “together, they leave little doubt about the celebrity the ad is meant to depict.”[6] Ms. White subsequently won a jury verdict on her claim.
Similarly, a court held that Ford Motors’ use of a voice that mimicked Bette Midler’s “distinctive” and “widely known” voice to sell cars violated her right of publicity. As the court noted, “[a] voice is as distinctive and personal as a face,” and that the voice is “one of the most palpable ways identity is manifested.”[7] In yet another context, a court held that a baseball pitcher’s stance was so distinctive that fans could identify him, therefore permitting the pitcher’s right of publicity claim to go before a jury.[8]
These examples help illustrate a couple points highly relevant to generative AI. First, the definition of “persona” is intentionally flexible. Bright-line rules such as “voices cannot constitute one’s persona,” generally do not exist. This ambiguity is the result of the numerous combinations of characteristics that can “imbue” one’s identity with economic value. In other words, what you are known for constitutes your public persona, which almost necessarily will consist of many component attributes. In the most rudimentary example, however, if you are a hand model, your persona includes your hands, not your voice; if you’re a singer, your persona likely includes your voice, likeness and musical style, but probably not your hands. Second, a central issue in these cases is whether the relevant segment of the public can identify the plaintiff. If the relevant demographic can identify the plaintiff’s persona from the commercial work, then it is likely an infringing use.
Cases involving the utilization of generative AI to create content that heavily borrows from an individual’s persona can seamlessly fit into this existing framework. When a company employs AI to mimic an individual's voice, likeness, or other indicia of personal identity for commercial purposes, courts can rely on abundant precedent to determine whether such use infringes upon the individual's right of publicity. The fact that the content in question was created using AI should not influence the ultimate outcome, as the focus remains on the infringement of the person’s rights rather than the method of content creation.
However, the proliferation of generative AI will undoubtedly result in a wave of new cases that stretch this jurisprudence. While earlier cases in this area were focused on companies using celebrities’ likeness in advertisements, generative AI democratizes the process, allowing anyone to create an image, song, or text. The myriad of uses that individuals will explore in both commercial and quasi-commercial contexts will present novel and difficult cases. Moreover, the extent of the “transformative” nature of AI-generated content will be contentiously debated (both in the publicity and copyright contexts). If a defendant’s use “transforms plaintiff’s identity to a sufficient degree,” then the defendant’s inclusion of the plaintiff in an expressive work may be immunized.[9] Finally, the determination of secondary and contributory liability for owners of generative AI models will play a pivotal role in shaping the rules and regulations governing this space.[10] Courts are already grappling with claims of copyright infringement involving OpenAI's products, and it is only a matter of time before defamation and right to publicity lawsuits arise.
(c) Closing remarks
Finally, while we discussed the parameters of the unlawful use of persons’ likenesses in connection with generative AI, there should be mention of increased opportunities for lawful uses of generative AI. Individuals’ right of publicity, like various other commercial rights, can be licensed and restricted.[11] With the increased ease of producing peoples’ likenesses, the opportunities for individuals whose likenesses have commercial value will only increase.
Through thoughtful regulation and as courts address these legal issues, our challenge as a society is to strike a delicate balance between protecting individual rights and enabling innovation.
---
[1] In general, music in the style of someone else is not considered a derivative work for purposes of copyright law and is allowed. But if the output is derived by machine learning and AI is generating the work based on the original copyright owner’s music with the input to create a song “in the style” of that artist, this may be an open question going forward.
[2] The Rights of Publicity and Privacy, McCarthy and Schechter, § 1:35 (2019 ed.).
[3] As of this writing, at least 36 states have recognized some form of the right of publicity through either statute or common law. No state has explicitly rejected the tort.
[4] The Restatement Second of Torts recognizes four privacy torts: intrusion, appropriation of name or likeness, unreasonable publicity, and false light. See Restatement (Second) Of Torts §§ 652A - 652I.
[5] To prevail under a right to publicity claim, the plaintiff must show: (1) the validity of the plaintiff’s right of publicity, and (2) that this right has been infringed upon by the defendant. To show infringement a plaintiff must generally show that the defendant, without permission, used some aspect of the plaintiff’s identity or persona in such a way that the plaintiff is identifiable from the defendant’s use; and the defendant’s use is likely to cause damage to the commercial value of that persona. This typically requires use for some commercial purpose.
[6] White v. Samsung Electronics America, Inc., 971 F.2d 1395.
[7] Midler v. Ford Motor Co., 849 F.2d 460.
[8] Newcombe v. Adolf Coors Co., 157 F.3d 686.
[9] The Rights of Publicity and Privacy, § 8:72.
[10] See The Rights of Publicity and Privacy, § 3:20 (discussing secondary liability under the right to publicity).
[11] The Rights of Publicity and Privacy, § 1:7.
