A new AI music model launches with fanfare, backed by licensing deals and partnership announcements. Days later, two of the world’s largest record labels file a lawsuit calling that same model “the fruit of the same poisoned tree.” This is not a contradiction. It is, in fact, the central tension now defining how copyright law will adapt to generative AI, and the Suno case is where that tension is playing out most visibly.
The Partnership That Did Not Buy Peace
Suno’s release of its v6 model came with a notable framing: the company described it as a way for AI and the music business to “move forward together.” Warner Music Group was among the partners involved. The announcement carried the implicit suggestion that Suno had turned a corner, moving from a company accused of unauthorized training toward one operating within a licensed framework.
Universal Music Group and Sony Music were not persuaded. Within days, they filed a new lawsuit, adding more than 61,000 songs to the legal dispute and arguing that v6 does not represent a clean break from the practices they had already challenged. Their core claim is that v6 was built, at least in part, on outputs and user interactions generated by earlier Suno models, a process sometimes called training on synthetic data. Suno’s own CEO, Mikey Shulman, mentioned those “interactions” in an interview with Billboard’s On The Record podcast when asked about the training data behind the new model.
The labels’ lawyers describe this as laundering rather than remedying the original problem. If the earlier models were trained on copyrighted recordings without authorization, then any model trained on the outputs of those earlier models carries the same legal contamination forward. The partnership deals, in this reading, do not reset the clock.
The Argument That Backfired
Here is what most coverage of this lawsuit misses. The legal stakes go beyond the question of whether Suno used copyrighted music without permission. They extend to a specific argument that Suno and other AI companies have relied on heavily in court: the fair use defense.
Fair use allows copyrighted material to be reused without payment in certain circumstances. One of the key factors courts weigh is whether the new use harms the market for the original work. Suno’s legal team had argued that no market for AI training data existed, because nobody had been paying for it. If there is no market to harm, the argument goes, there is no fair use problem.
The new lawsuit turns that argument against Suno directly. By entering into licensing agreements with music companies, including Warner, Suno has now become what UMG and Sony describe as “a repeat, paying participant” in exactly the kind of market it previously claimed did not exist. The labels argue that Suno can no longer credibly deny that such a market exists, because its own business conduct proves otherwise. This is a significant legal development, and it did not require the labels to introduce new evidence. Suno’s own deals provided the ammunition.
Suno, for its part, maintains that the lawsuit’s claims are “fundamentally flawed on both the facts and the law,” and that v6 was trained on licensed content, community interactions, and accumulated learnings from its team.
Why This Case Reaches Beyond Music
The Suno dispute is one instance of a much broader legal confrontation. Authors, newspapers, visual artists, film studios, and record labels have all brought copyright claims against AI companies over training practices. The underlying question is the same in each case: does using millions of existing works to train a generative model constitute infringement, or does it fall within fair use as a transformative act?
That question remains unresolved in courts. The stakes, as the source material notes, are described as carrying trillion-dollar implications for both AI development and creative industries.
What the Suno case adds to this broader picture is a specific argument about what happens when AI companies begin licensing content after years of operating without licenses. Does that shift their legal exposure retroactively? Does it undermine defenses they have already filed? The labels say yes. Suno says no. Courts will eventually decide.
The deeper issue, articulated in the lawsuit itself, concerns the relationship between human creative labor and the systems built on top of it. The labels’ lawyers note that Suno’s CEO has publicly dismissed the time, practice, and skill involved in making music, while the company’s product depends entirely on recordings that are the result of exactly that labor. Whether or not that observation carries legal weight, it captures something real about the structural tension at the heart of generative AI: these systems are trained on human output, and the humans who produced that output did not necessarily consent to that use.
In Short
Suno released a new AI music model with licensing partnerships and positioned it as a fresh start. UMG and Sony responded with a new lawsuit adding more than 61,000 songs to the dispute, arguing the new model was built on the outputs of earlier, allegedly infringing models. Crucially, they also argue that Suno’s own licensing deals have destroyed its central legal defense, since those deals prove a market for AI training data exists. The case is one of the clearest examples yet of how generative AI’s legal reckoning is not waiting for technology to mature. It is already here.
Based on reporting from Billboard - AI.