September 2026
Suno has always been a pain in the ass for the music industry, and only the laziest person didn’t attack them. Some argued that proper attribution and revenue sharing could be a feasible solution for music created with heavy use of AI. And the latest Suno update, called v6, is the first attempt to implement this in practice.
While recent announcements over the past few days shed some light on how it is going to work (licensed data from WMG, Believe and BMG, as well as revenue sharing, are included) some fundamental questions remain open for me.
How do they deal with the complexity of music rights, which are not limited to sound recordings?
In an interview with Billboard, Suno’s CEO admitted that this is a problem and shared his belief in a bright future where it becomes less of an issue thanks to further technological developments. For now, he said they would start with a limited volume of music that is easier to clear for licensing. I assume this is going to have a tremendous impact on the size of the dataset and potentially change the quality of the outputs. Some people have already started sharing their observations, but I’ll come back to that later.
How do you track the revenue to be shared if music created on Suno is downloaded and distributed elsewhere?
Here, things are less obvious.
The CEO of WMG, in his memo, does not mention participation in distribution revenue. He focuses primarily on the monetisation of music creation. What happens next, after creation, if and when the music leaves Suno’s garden, remains ambiguous.
At the same time, Believe’s announcement clearly states that they will remove their previous ban on AI music and make music created with Suno eligible for distribution through Believe. I can only wonder how BMG and Warner feel about it. Does this create an inherent conflict that all AI companies involved in music creation will eventually encounter? If data is licensed by multiple copyright holders, who gets the right to distribute the outputs?
Does the new model really use only licensed music?
In the same Billboard interview, host Kristin Robinson brilliantly insists on clarifying what data is going to be used in v6. The answer was a bit vague and, to me, potentially confusing. On the one hand, Mikey Shulman admitted that no unlicensed third-party data is being used. On the other, he referred to Suno users’ data, which I interpret as synthetic original data — music created by users using previous models that relied heavily on unlicensed music.
My understanding is that Suno is trying to avoid expanding the scope of its potential infringement by not directly using the original music it previously ripped from YouTube.
Another question arises around Suno’s challenge to the YouTube “stream ripping” claim brought by UMG and Sony. If Suno considers stream ripping to be fair use, does that imply that anyone stream-ripping songs made by Suno’s users would also not violate anyone’s rights? It sounds to me like shooting yourself in the foot.
Finally, we already have some early reactions from Suno users to the new v6 model. With all the limitations that come with a smaller dataset, it is reasonable to expect the quality to be different from what it was before. One of the most-liked comments on YouTube under the Billboard interview says: “it sucks ....the sound quality has gotten worse than version v 5”
But there is another one that is even more interesting to me: “Beyond that, the music itself often sounds generic, indecisive, hollow, and emotionally flat.”
Suddenly, the model that primarily uses synthetic data is losing its soul.