Is Your Favorite New Artist Actually an AI Robot?

We take action against licensors and distributors who fail to police for this kind of fraud, and those who commit repeated or egregious violations can and have been permanently removed from Spotify. SoundOn is owned by ByteDance, the Chinese company that also owns TikTok.

The incident highlights an inherent problem with AI music uploads: There are no filters, no clear guardrails, and no disclosure that any of the content on Spotify is AI-generated. Apple Music and Amazon Music also don’t label AI content. YouTube labels all AI tracks created using its own generative AI tools and requires disclosures for “Synthetically generating music.” Creators are encouraged to voluntarily label their tracks, but YouTube may apply the label for them if it deems necessary. All of the big streamers require submissions to be owned by rights holders; they all say there’s zero tolerance for impersonators or copyright infringement.

These incidents aren’t flukes. In 2023, a TikTok user uploaded an AI-generated track called “Heart on My Sleeve” to the social media app and music streaming platforms, claiming to be the work of music artists Drake and the Weeknd. Before the “Fake Drake” track was pulled by Universal Music Group for copyright infringement, it generated a reported 600,000 streams on Spotify, 5 million views on TikTok, and 275,000 on YouTube.

Shortly after the Blaze Foley incident, New York City-based saxophonist Chris Ward told me he saw entire albums attributed to a John Scofield impersonator. Unlike Foley or Guy Clark, the prolific guitarist John Scofield is alive and well. The albums credited to him, like “Background Guitarra Jazz,” were clearly designed to feed playlist algorithms. When I notified Spotify of the Scofield imitations, the content was quickly removed — only to reappear weeks later. But there are dozens of these instances being cataloged by Reddit users, across all genres.

“I’ve seen cloned AI versions of jazz records on YouTube and Spotify — stuff released under real musicians’ names that they never recorded,” says Ward. “Sometimes it even carries the liner notes of the original album. It’s fake revenue, fake art, and no one knows where the money goes.”

Eric Drott, a music theory professor at the University of Texas at Austin, says that Spotify and other music streaming platforms limit their liability by requiring uploads to go through distributors. “It creates layers of mediation where bad actors can insert themselves,” says Drott. “And because digital reproduction is so cheap — and now purely AI-generated music is cheap — you can imagine scammers saying, ‘This will get taken down eventually, but can we rack up enough streams in the meantime to cash out?’ Multiply that by hundreds of obscure artists, and those numbers start to become significant.”

After the “Fake Drake” track was pulled from streaming platforms, UMG — which holds a financial stake in multiple platforms — warned that AI-generated music puts the industry at a crossroads. The label said this “begs the question as to which side of history all stakeholders in the music ecosystem want to be on: the side of artists, fans, and human creative expression, or on the side of deep fakes, fraud, and denying artists their due compensation.”

Jael Holzman, vocalist, lyricist, and bassist of Ekko Astral — a punk band from Washington, D.C., whose debut album “pink balloons” was named Pitchfork’s record of the year in 2024 — recounted learning about shady AI-generated tracks masquerading as new music by deceased rappers like Nipsey Hussle and Pop Smoke. The tracks appeared to originate from someone in upstate New York who used AI to create fake rapper personas, complete with multiple alter egos and impersonations of real artists. In one case, rapper Riff Raff’s likeness and verse were used without permission, confirmed by the rapper’s management, says Holzman, who is also a journalist.

Musical creators are protected under the U.S. Copyright Act of 1976, but the courts haven’t addressed how AI-generated content should be treated, in particular when programs are trained on an artist’s protected works.

In a paper titled “Fake Drake? AI Music Generation Implicates Copyright and the Right of Publicity” by Hope Juzon for the Washington Law Review, the author writes, “If courts hold that the use of copyrighted inputs constitutes infringement, there may be a chilling effect on AI music generation as a whole. Alternatively, if courts find that this use is not infringing, rights holders may lose control over their works and fail to receive compensation for the use of their works.”

Other creative industries face parallel challenges of AI programs being trained on their work, including publishing, visual art, film, and digital media. On Sept. 5, Anthropic, the AI company behind the large language model (LLM) Claude, agreed to pay a $1.5 billion settlement in a lawsuit brought by a group of authors and publishers.

The judge ruled that Anthropic had illegally downloaded millions of books to train its LLM. But the ruling wasn’t entirely one-sided: the judge held that Anthropic’s use of legally acquired books to train on AI can qualify as fair use. The ruling could set a legal precedent for how copyright cases are handled across other creative industries, including music.

Spotify is already a tough gig for most musicians. Emerging after the wild west days of Napster and Kazaa, legal streaming was a revelation. No longer did you have to blow your budget on individual records or wait days to illegally download an artist’s entire catalog. For roughly the price of a single album, you can access a massive library of music on demand. The rise of on-demand audio streaming has seen significant growth in recent years, with total plays and paid subscriptions increasing annually. In 2024, on-demand audio streams in the U.S. reached approximately 1.4 trillion, a 6.5% increase from the previous year, according to Luminate. Additionally, paid streaming subscriptions in the U.S. rose to 100 million in 2024, up from 75.5 million in 2020, as reported by the Recording Industry Association of America (RIAA).

While consumers benefit from the value proposition offered by music streaming platforms, the same cannot be said for artists. Most artists earn very little from streaming due to the complex revenue calculation systems and existing royalties structure in place. Platforms like Spotify pay artists as little as $0.003-$0.005 per stream, requiring millions of streams for artists to earn a substantial income, further divided among various rights holders.

Despite Spotify paying $10 billion to the music industry in 2024, global recorded music revenue reaching $30 billion, and the increase in the number of artists earning meaningful income, many artists still struggle to make a living from streaming. Low-visibility artists face a dilemma of relying on algorithms for visibility or accepting lower royalties through tools like Spotify’s Discovery Mode.

The introduction of AI playlists by Spotify CEO Daniel Ek poses additional challenges for human artists. AI-generated content, including tracks from anonymous session musicians, has been used to populate playlists, reducing royalty payouts for traditional artists. This trend could potentially lead to a future where robots produce endless tracks for playlists, further depriving human artists of their fair share of streaming revenue. Independent artists are at a disadvantage compared to major label artists because they lack resources. Spotify’s minimum thresholds could harm them by preventing them from reaching the payout threshold if an AI track draws enough listeners away from them. This could essentially demonetize them.

Comparison of Platforms’ Approach to AI-Generated Music

sentence: Please make sure to lock the door before you leave. statement to improve clarity:

Please rewrite the statement to enhance its clarity.