Universal and Sony went back to court against Suno on 18 September, in Boston, and Universal filed a separate case against distribution platform DistroKid three days before that, in Delaware. If you’re not a lawyer and don’t plan on becoming one, it’s easy to file that under “labels being labels” and move on. But both cases are really about the same question every working musician, producer and session player has been quietly sitting with all year: when a track has AI somewhere in its signal chain, who gets paid, who owns it, and who’s liable if it turns out to be built on someone else’s work?
The Suno case: what changed with v6
Universal and Sony’s first suit against Suno, filed back in June 2024, covered 560 recordings and argued the AI platform’s models were trained on copyrighted music without permission. Suno responded the way a lot of AI platforms have this cycle: it cut licensing deals with labels who might otherwise sue. Warner signed on in November 2025, BMG followed in August this year, and independent-rights network Believe joined on 8 September.
Universal and Sony haven’t signed, and on 9 September, Suno launched v6, which the company’s Chief Product Officer Jack Brody says was “trained entirely from scratch” without Universal or Sony’s catalogues. The new suit, covering 60,202 recordings, argues that doesn’t matter: it claims v6 was built using preference data, knowledge distillation and retained copies drawn from the earlier, allegedly infringing models. As the labels put it in their filing, “training a ‘new’ model on the outputs of an infringing model does not eliminate the infringement; it launders it.” Suno’s position, in Brody’s words, is that the labels’ own partnership revenue shares were “not in exchange for training” and “not really about the data” in the first place. Nobody’s proven anything yet, it’s a genuine, unresolved dispute over provenance, and it’s the fight worth watching, because if the labels win the “laundering” argument, it becomes much harder for any AI platform to simply retrain its way out of a lawsuit.
The DistroKid case: distribution, not just generation
The DistroKid suit is a different animal. Universal isn’t arguing DistroKid built an AI model, it’s arguing DistroKid built the pipeline that gets AI tracks onto streaming platforms while looking like ordinary artist releases, then kept distributing tracks after being told it didn’t have the rights to them. The numbers UMG cites are the real headline here: of 1,551 AI-made tracks that hit charting service SIQA’s radar in the first quarter of 2026, 90.4% were made with Suno and 75.8% went out through DistroKid. That’s not a coincidence, it’s a supply chain, and it’s the first time a major has gone after the distribution layer rather than just the generation layer.
What’s actually settled, and what it means day to day
Here’s the part that matters if you’re not a lawyer or a label executive. Away from the courtroom, a working system has quietly formed around the platforms that did settle. Suno, Udio and others that struck licensing deals now pay the majors and networks like Merlin for training access, and in return, paying subscribers on those platforms get two things: the right to distribute what they make commercially, on Spotify and elsewhere, and indemnification against the kind of “sounds suspiciously like someone else’s record” dispute that would otherwise land on the user, not the platform. Streaming services aren’t blanket-banning AI tracks either; Spotify and Apple Music are leaning on detection tools, including Believe’s AI Radar, to flag undisclosed AI content while letting through anything that’s properly licensed and labelled.
Ownership is the bit still catching people out. The US Copyright Office’s position hasn’t moved: a purely AI-generated file, prompt in, track out, can’t be copyrighted. What’s opened up is a “hybrid workflow” path, where adding meaningful human authorship, re-recording a vocal, replacing an AI drum part with a real performance, swapping in your own mix, turns the result into something you can actually register with ASCAP, BMI or a PRO of your choice. If you’re using AI tools anywhere in a track you intend to release and monetise, that’s the line worth knowing: label what’s AI, keep a human hand in the final version, and get it in writing, a split sheet, same as any other collaboration, if more than one person touched it.
None of this changes what’s always been true at the affordable end of the market for session players and producers either. Automated mastering is already genuinely competitive on price for streaming-first genres, free through BandLab, a few dollars through Waves or LANDR, which puts more pressure on human engineers to be worth the premium through taste and relationship, not just turnaround time. The lawsuits are about the platforms. The mastering chain is about you. Both are worth watching the same week.
0 Comments