Kai Noir loves beats. The DJ produces deep tech house and is constantly working on new tracks. He already has several thousand monthly listeners on the streaming platform Spotify. But there's a surprise: Kai Noir doesn't exist. At least not as a flesh-and-blood human being. He is an AI avatar created by Daniel Troha.
The music producer and entrepreneur says he founded Germany's first label for AI music. His songs are primarily created using the AI software “Suno”. A short prompt is enough and the program creates a completely new music track in just a few minutes.
At the beginning of 2026, the US company “Suno” reported two million paying subscribers. Daniel Troha also uses such software. But professionals like Troha continue to work on the AI songs, for example fine-tuning vocal passages or adding guitar sounds or other instruments. The machine starts, the human complements, changes and refines. Troha calls the AI tool a “studio production buddy” that saves him “an incredible amount of time and work.”
Sold-out stadiums, real musicians playing real instruments - it is precisely this established system in the music industry that is being challenged by artificial intelligence. There are now countless AI songs on various streaming platforms.
Some give the impression of being sung by people. Like the neo-soul singer Sienna Rose, who at times had more than four million monthly listeners on Spotify alone. Rose even had a presence on social media.
However, research by Bayerischer Rundfunk suggests that there is an AI project behind Rose and that there is no real singer with that name. The band “The Velvet Sundown,” which has been streamed millions of times, is also AI.
This is possible because AI platforms like “Suno” have been trained with millions of songs. The process: Existing works by musicians are analyzed and converted into training data. AI music is only possible through this human input. These also include well-known titles from Germany. For example “Breathless through the Night”, “Forever Young” or “Daddy Cool”.
GEMA protects the copyrights of music creators. The collecting society accuses “Suno” of using songs without permission and without payment. GEMA legal advisor Kai Welp criticizes: "Ultimately, an unimaginable number of musical works were taken just like that." Money is then made with this training data without asking the creative people. For the general counsel, this is “a very perfidious business model”.
GEMA therefore took legal action against this practice and was successful at the Munich I Regional Court at the end of July this year. According to the ruling, “Suno” may not reproduce six well-known works from the GEMA repertoire without a license or use them to train its AI. The company must also provide information and pay compensation, the amount of which is still unclear.
“Suno” admitted that it had used the works for training purposes, but denied any obligation to pay remuneration. The company said it disagreed with the ruling and was considering an appeal. GEMA hopes that such judgments will in future oblige AI music platforms to give musicians a share in their sales.
A study commissioned by GEMA and its French sister company SACEM comes to a drastic conclusion. According to this, around 27 percent of the income of music creators in Germany and France could be at risk from generative AI by 2028.
Looking over five years, the study published at the beginning of 2024 puts the possible damage at around 2.7 billion euros. According to the study, seven out of ten musicians surveyed also see their economic basis threatened by AI.
Whether AI music and fair pay for music creators can be reconciled depends largely on the streaming platforms. The more often a song is played there, the more money the rights holders receive from the shared pot.
Upon request, market leader Spotify will refer in writing to protective measures against fake songs. AI music must be consciously selected by users. It will also be marked if a profile depicts an AI artist.
Competitor Deezer, with almost nine million subscribers, uses its own software to recognize AI songs. These would be flagged and excluded from algorithmic recommendations by default. “We label AI songs so that users can know what they are listening to,” says Deezer boss Alexis Lanternier. Anyone who wants to listen to AI music can do so. “It should be a conscious decision on the part of the user,” Lanternier continued.
The quantity is growing rapidly. In January 2025, the share of AI songs in Deezer's daily uploads was still around ten percent, around 10,000 titles per day. In November 2025 it was already 34 percent, i.e. around 50,000 songs. In June 2026, an average of around 90,000 AI songs were uploaded every day - more than half of all new titles.
For some, AI songs are just another tool, for others they are a threat to the entire music industry. It remains to be seen what a system could look like that fairly involves musicians - and shows listeners who they are actually listening to.
