Introduction
As we discussed previously, AI can have serious consequences for intellectual property (IP) law, and it is reasonable to assume that intellectual property is the area of law most affected by AI. One of the first big issues recently is songs using AI-generated vocals. These look like famous artists (like Drake, The Weeknd, Rhianna, etc.) or characters (like SpongeBob, Eric Cartman, Goku, etc.).
The story begins with the song “Heart on My Sleeve”. The song is, to our knowledge, an original. But it features AI-generated vocals that sound like those of Drake and The Weeknd. Drake and UMG (Drake’s label) weren’t happy about it. However, since the representatives are not protected by copyright, they did not have the option to remove the song from YouTube.
This is a new problem; no case law can therefore help resolve this situation. However, this isn’t the first time artists have had to deal with someone else using their voice; they are not entirely defenseless. To what extent can existing case law or legal principles be helpful?
In some US states, artists may have the ability to sue the creators of these AI vocal songs based on the “right of publicity”, a term established in law. Haelan Laboratories, Inc. v. Topps Chewing Gum, Inc., which protects the images of celebrities. On the other hand, artists can still count on unfair competition in other countries.
Analogous case law on AI-generated voices
In this case, Ford attempted to hire a famous singer, Bette Midler, for a commercial. Once Midler refused, Ford instead hired one of his backup singers, Ula Hedwig, to sing Midler’s song and imitate her voice. Notably, Ford obtained a license from the copyright holder to use the music. Therefore, the central issue in this case was solely the protection of Midler’s voice.
The Court made several important conclusions in this case.
First, the Court held that the First Amendment (right to free speech) protects the reproduction of images and sound if the purpose of the copying is informational or cultural. However, if the reproduction is intended to exploit the individual depicted, then the First Amendment will not apply. Likewise, in Apple Corp v. Leber et al. And Estate of Presley vs. Russencourts have held that the First Amendment does not protect imitative entertainment without creative elements.
Second, the Court stated that imitation of a recorded performance does not constitute copyright infringement, even when one artist deliberately attempts to reproduce another’s performance.
Finally, the Court concluded that harm can result from the appropriation of attributes of a person’s identity. Although a person’s voice is not protected by copyright, the Court determined that a person’s voice is as distinctive and personal as their face. When the unique voice of a professional singer is widely known and is intentionally imitated in commercial use, the sellers have misappropriated what does not belong to them and have committed a crime in California.
Tom Waits was an acclaimed singer with a deep, distinctive voice. Frito-Lay, Inc. was a large manufacturer and distributor of food products. To launch a new corn chip salsa flavor, Frito-Lay (their advertising agency, to be specific) hired Stephen Carter, a professional musician who had spent years covering Waits songs, to record a jingle for advertising. After the ad aired, Waits sued Frito-Lay and the ad agency.
This case upheld the Court’s findings in Midler that “where voice is a sufficient clue to a celebrity’s identity, the right of publicity protects him or her from being imitated for commercial purposes without consent.” of fame” and clarified the common law rule that for a voice to be misappropriated, it must be (1) distinctive, (2) widely known, and (3) deliberately imitated for commercial purposes.
Additionally, the jury reasoned that because the voices of Stephen Carter and Tom Waits are similar, they could mislead consumers into believing that Waits endorsed the corn chips. This is entitled Expectation of damages under the Lanham law (due to unfair competition).
AI-generated voices in the Western Balkans
As we can see, this problem was present (without AI) in the United States. However, a similar case in Serbia involves the creation of AI-generated media content using the voices of well-known politicians. The absence of other legal remedies allowed the problem to be resolved on a regulatory basis. The Electronic Media Regulatory Authority of the Republic of Serbia (REM) has warned media service providers. According to the Electronic Media Law, providers must not broadcast content that could exploit the credulity of viewers and listeners. Furthermore, audiovisual content must not mislead the public regarding identities, events or events. Failure to comply with the above may result in license suspension or revocation. Here we can see similarities between Serbian and American laws. Misleading the public was the basis for the Lanham Act damages award in Waits v. Frito-Lay, Inc.
REM further stated that AI content should have explicit notification before, during and at the end. It should show that the program uses AI. Additionally, media service providers must prevent any misuse of personal data or identities, including voice or image.
Exit
Ultimately, we can see that artists and record labels are not completely helpless against AI-generated voices that imitate famous artists. Where intellectual property laws do not protect against intellectual property infringement, unfair competition rules and protection against misleading content do. Some US states allow artists to assert their right to publicity, while in some countries, such as Serbia, relevant agencies can revoke a broadcast license if AI is used to mislead the public. Almost all countries recognize the right of artists to invoke unfair competition if there is a real risk of misleading consumers. The fight in Serbia could be more difficult. Protection against unfair competition is provided to a limited extent by the Trade Act.
But that doesn’t solve this problem for YouTube (and other platforms), because they need an effective platform-wide solution. This is why YouTube is in talks with UMG to resolve this issue.
A possible solution to this problem would be, for example, for YouTube to allow users to prevent the AI from practicing to their songs. Other companies are also following this trend, including Spotify and Zoom, which recently changed their terms of service so users can opt out of having their data used to train AI. Another potential solution could be to pay royalties every time an artist’s voice is used to create a song.