A recent viral song that used artificial intelligence-generated vocals of Drake and The Weeknd sparked outcry and intrigue in an industry already leery of unregulated AI music, which has driven a wedge through multiple intellectual property rights.
The song “Heart on My Sleeve,” published by an unknown TikTok user who goes by “ghostwriter977,” was also released last month on music streaming platforms like Spotify, Apple Music, and YouTube, racking up millions of views and publicity over its catchy lyrics. But
UMG had already been urging the platforms to stop AI companies from using its tracks to train generative music programs. The National Music Publishers Association sent a letter to US Senate Majority Leader
Musicians and record labels may have a few legal avenues to combat unauthorized AI generated music, including state-based right of publicity laws and federal copyright law. But those theories have yet to be tested in court, and AI companies and users will likely invoke the First Amendment and other defenses.
“That’s where we have ourselves right now, at this intersection where legislation hasn’t passed because it’s all brand new,” said Kendall Minter, an entertainment attorney at Greenspoon Marder LLP. “It’s all cobbling together based on what I’m sure will be upcoming litigation.”
Right of Publicity
Drake, The Weeknd, and other artists’ best bet to stop AI-generated vocal imitators under right of publicity laws—which generally protect one’s name, image, and likeness from being exploited without permission—attorneys said.
But unlike copyright and trademark law, the right of publicity is based entirely on state laws. Some like New York and California provide explicit statutory rights, while others are based only in common law.
“Whether a particular state’s right of publicity law is going to cover something like somebody’s voice is going to depend on the state,” said Scott Sholder, an intellectual property attorney at Cowan DeBaets Abrahams & Sheppard LLP.
California’s law does explicitly cover voice, which the US Court of Appeals for the Ninth Circuit clarified in the 1988 Midler v. Ford Motor Co. decision.
In that case, singer Bette Midler sued the automaker for releasing a television advertisement that used an actor to imitate her voice and sing a song from one of Midler’s albums. The court determined that California’s statute didn’t cover voice imitations because Ford was using the actor’s voice in the ad, not Midler’s. However, the court said that Midler did have common law rights to her voice that would allow her to advance the lawsuit.
But other states’ courts may come to different conclusions based on their laws, potentially creating a disparate web of rulings across the country.
The fact that the AI-generated vocals are being used to create music and not commercial ads or products could complicate a right of publicity claim. A potential defendant could argue that the First Amendment protects AI vocals in an artistic work like a song.
Courts have long used a free speech balancing test—the Rogers test—that asks if the unauthorized use of IP is artistically relevant to the expression and not explicitly misleading to consumers. While often used for trademark infringement cases, the test originated in Rogers v. Grimaldi, a 1989 Second Circuit decision that shot down actress Ginger Rogers’ right of publicity and false endorsement lawsuit.
“The plaintiff’s side will say, no this is actually a commercial product because you’re selling downloads of the song or you’re getting paid per stream on Spotify,” Sholder said. “That’s going to be a thorny question.”
What About Copyright?
UMG requested “Heart on My Sleeve” be removed from the streaming platforms using copyright take-down requests. The song contains a short audio sample of the well-known tag of producer Metro Boomin, which does have copyright protection.
But without that tag, UMG likely wouldn’t have any direct copyright claims to an otherwise original song with original lyrics. Courts have held that voice or musical style can’t receive copyright protection.
It remains unresolved whether AI programs—like those that can produce voice imitations or new melodies—violate copyright law by training on millions of copyrighted songs and sound recordings without a license.
Companies behind generative AI art programs are facing different lawsuits brought by artists, open source coders, and Getty Images over machine-learning algorithms that swept up their content.
Whether those like ghostwriter977 who use AI tools to create music can receive copyright protection is also unclear.
The US Copyright Office has denied registering copyrights for works created entirely by AI, but it has been more open to offering protections to works that were created with the help of AI.
“In the Copyright Office’s view, it’s a bit like the photographer: Are you just pushing the button, or are you adding other inputs like framing the photo,” said Shubha Ghosh, an intellectual property professor at Syracuse University.
—with assistance from Kyle Jahner
To contact the reporter on this story:
To contact the editors responsible for this story: