Musical Heist: The dangers of AI music and the assault on transparency, consent and remuneration

Last week, a song which has been streamed millions of times in Sweden was banned from the country's music charts because it was created by Artificial Intelligence (AI).

Called I know, You're Not Mine (Jag vet, du är inte min) and which went on to top the Spotify playlist of Sweden's most popular songs, the hit is a folk-pop song telling a melancholic story of lost love. 

But the singer is a digital creation and the country's music industry body officially blocked the track from its official chart listings.

Quickly, the song, attributed to a singer called 'Jacub', became Sweden's biggest song of 2026 so far, amassing more than five million Spotify streams in a matter of weeks, putting it at the top of the platform's Swedish Top 50.

However, journalists who began investigating Jacub's identity found that the artist had no significant social media profile, media appearances or tour dates.

When investigative journalist Emanuel Karlsten began digging deeper, he found that the song was registered to a group of executives connected to Stellar Music, a music publishing and marketing firm based in Denmark. Two of the individuals work in Stellar's AI department.

The producers – calling themselves Team Jacub – issued a lengthy email to the media, insisting their creative process had been misunderstood.

"The team behind Jacub consists of experienced music creators, songwriters, and producers who have invested a lot of time, care, emotions, and financial resources," the email read in part.

They described AI as a "tool" or an "assisting instrument" within a "human-controlled creative process". To Team Jacub, they said, the five million Spotify streams were proof of the song's "long-term artistic value."

Training AI models often requires copying entire works or substantial portions into datasets, transforming them into machine-readable formats, and creating temporary or intermediate copies throughout the training process. 

Rights holders contend this constitutes copyright infringement unless a statutory defense, such as fair use, applies.

Papaoutai.

Even before the year began, on December 20th, a new song appeared in Spotify's catalogue. 

It had already been circulating informally on TikTok since at least the beginning of the month. Given its growing popularity on the Chinese social network, the track was released on the Swedish platform by an "independent" label, also from Sweden, called Unjaps.

The music is familiar to music lovers, as it's a cover of Stromae's popular hit song "Papaoutai." 

The unique aspect of this version is that no human actually sings; it's entirely generated by artificial intelligence (AI). 

Since the beginning of January, the song has been a resounding success. It has climbed into Spotify's top 100 most-streamed tracks, achieving notable popularity in several countries, including the United States. 

On YouTube, accounts have been covering this “Afro Soul” version of “Papaoutai,” racking up millions of views, and on Instagram and TikTok, hundreds of thousands of videos also feature this cover as background music.

AI-generated output may cross into unauthorized derivative work when it closely replicates the structure, cadence, melodies, or organization of copyrighted work, such that it is substantially similar, meaning that even temporary duplication of copyrighted text or audio can trigger infringement concerns. 

For example, Xania Monet, an AI-driven artist, has charted on Billboard and attracted a $3million record deal.

Taken together, these developments underscore that AI-associated music is increasingly competing with human-created works for the same deals, chart recognition, and audience attention.

In the case of “Papaoutai,” Belgian public television noted a striking resemblance between the sound and an atmosphere reminiscent of The Lion King. The song's artwork also depicts a savannah, often with a sunset, much like the Disney animated film. 

Did the creators of this version ask AI to create a blend of The Lion King, Stromae, and an African-inspired sound? No information has been released on this subject.

Copyright proofing.

The worry is that AI will eventually absorb all creative works in history and spew out endless slop that will replace human-made art and drive artists into penury.

Those worries are being deepened by how the major labels, once fearful of the technology, are now embracing it – and heralding a future in which ordinary listeners have a hand in co-creating music with their favourite musicians.

AI music platforms analyse huge amounts of recorded music in order to learn its sounds, structures and expressions, and then allow users to create their own AI-generated music via text or speech prompts.

You might ask for a moody R&B song about a breakup sung by a female vocalist, and it will come up with a decent approximation of one, because it’s absorbed hundreds of such songs.

According to Bloomberg Law, Artists and authors should review recording, publishing, and distribution agreements for clauses on new technologies, derivatives, text and data mining, machine learning training, voice likeness/right of publicity, and any clauses that let labels or publishers license on their behalf. 

Otherwise, though their rights may be clear today on one medium, such as streaming, their rights may be muddled, or non-existent on yet-to-be-developed mediums.

Aware of the scale of the shift, last year the Recording Industry Association of America, representing the three major labels, initiated legal action against AI music companies Suno and Udio for copyright infringement, alleging they had trained their AI platforms on the labels’ artists without their permission. 

But then there was an extraordinary about-turn - they didn’t just settle the matter out of court – Universal Music Group (UMG) then partnered with Udio, and Warner Music Group (WMG) with Udio and Suno.

Irving Azoff, artist manager and founder of the Music Artists Coalition in the US, responded to the Universal/Udio deal with biting cynicism. “We’ve seen this before – everyone talks about ‘partnership,’ but artists end up on the sidelines with scraps,” he said. 

In the wake of the same deal, the Council of Music Makers in the UK accused the major labels of “spin” and called for a more robust set of artist-label agreements. 

And the European Composer and Songwriter Alliance also reacted, saying there was a disturbing “lack of transparency” around the deals.

"The world's largest tech companies, as well as AI-specific companies, such as OpenAI, Suno, Udio, Mistral, etc., are engaged in the largest copyright infringement exercise that has been seen," John Phelan, director general of the International Confederation of Music Publishers (ICMP), told AFP.

He added: "What is legal or illegal is how the technologies are used. That means the corporate decisions made by the chief executives of companies matter immensely and should comply with the law and what we are seeing is that they are engaged in wilful, commercial-scale copyright infringement."

AI-generated music accounts for 28% of content uploaded daily on Deezer, the French music platform, which has reported "a surge" over the past year in uploads.

In April 2024, artists including Billie Eilish, Nicki Minaj, Katy Perry, Stevie Wonder and hundreds more signed an open letter urging AI companies to stop training their models on copyrighted music without permission.

Writing in the New York State Bar Association, Temidayo Akinjisola said: "In this moment, we observed that through the deal-making, judicial and legislative processes, we are beginning to see a framework develop that may soon require express authorized permission from rights holders for the use of their copyrighted material for AI training purposes."

"We are also observing that there is a pathway forward for works created with AI, but not without nuances and limitations."