Sony Music and UMG say Suno's new models still violates their copyright

The music industry's legal strategy is becoming clear: chase every checkpoint, every version, every "improvement" to the model. The Engadget report on the lawsuit shows that labels are not just litigating past behavior—they are trying to freeze generative AI companies in place by forcing each new release to prove it was trained cleanly. Suno's v6 may have found a roundabout method, but the labels' argument is blunt: if the training data was unlicensed, the model is tainted, no matter how clever the ingestion process became. This is the core tension of the AI copyright era. Courts are being asked to decide whether copyright infringement is about the substance of the training data or the mechanics of access. Suno will likely argue that its new models were trained differently, perhaps using synthetic data or some other intermediary. But the labels are preemptively rejecting that distinction, insisting that a workaround is still a violation when the underlying source material was never licensed. That framing matters because it shifts the burden onto AI companies to prove their entire lineage is clean. The broader implication is that generative music is now an arms race between legal exposure and technical evasion. Every time a company like Suno updates its architecture, it invites a new complaint. That creates a chilling effect: even if a model is trained on entirely original or licensed data, the mere existence of prior infringement claims makes future releases legally radioactive. The labels are not just suing for damages; they are trying to make the cost of iterating so high that AI music companies either capitulate or collapse. The source article (https://www.engadget.com/2262978/sony-music-and-umg-say-sunos-new-models-still-violates-their-copyright/) captures the immediate dispute, but the real story is about precedent. If the courts accept the labels' logic, then any future model that even tangentially builds on a tainted predecessor could be dragged into court. That would effectively mean the music industry can veto AI innovation by proving one bad training run. It is a high-stakes test of whether copyright law can keep pace with generative systems that evolve faster than litigation.