Suno's v6 Swaps Scraped Data for Licensing Deals With Warner, BMG

Suno just shipped v6, its first model family built with direct participation from the record industry rather than in spite of it. Chief product officer Jack Brody told The Verge that v6 was "trained from the ground up, with a new set of data that does not include the same data that our previous models were trained on." The new dataset draws on content licensed from Warner Music Group and BMG, plus user data from Suno's own platform.

That framing only makes sense against Suno's legal backstory. Major labels — Sony Music, Universal Music Group, and Warner — sued Suno in 2024 over alleged use of copyrighted recordings in training. Warner settled in November 2025, becoming the first major to do so, and BMG followed with a licensing partnership in August 2026, a month before v6 launched, according to TechCrunch. Believe is also part of the licensing group. Sony and UMG have not settled, and separate suits — including one from artist Jason Isbell and a user class action over data handling — are still active.

Three models, one pitch: legitimacy

The v6 family ships in three tiers: a base v6 aimed at reliable, steerable output for paying users; v6-wild, an experimental mode for paid users chasing unexpected results; and a free v6-mini for faster generation. New capabilities include prompt- or lyric-based editing of song sections, using text, images, or video as creative references, and instrument separation for remixing — alongside a previously announced watermarking system and a planned opt-in remix program with participating artists.

Brody described the goal as creating "additional revenue streams" for rights holders through derivative works, though he declined to disclose the financial terms of the label deals.

Why builders should care beyond the music angle

Suno isn't a fintech or trading story, but the pattern is one every AI builder should recognize: the market is pricing in the difference between a model trained on data you can defend and one you can't. Suno spent two years fighting suits from Sony and UMG while quietly negotiating licenses with Warner and BMG — and only after settling did it rebuild its flagship model on that licensed foundation, discarding the older training data entirely rather than patching around it.

For anyone shipping AI features on top of scraped or ambiguously-sourced data — training sets, fine-tuning corpora, RAG indexes pulled from the open web — the lesson generalizes past music: provenance isn't a compliance afterthought, it's a product decision that eventually forces a rebuild if you get it wrong. That's the same discipline we push in AI-assisted development work: know where your data and dependencies actually come from before you ship on top of them, not after a subpoena.

What's still unresolved

v6 doesn't make Suno's legal exposure disappear. Sony and Universal remain unsettled, artist-level suits are ongoing, and the company has separately acknowledged training earlier models on YouTube audio — a detail that keeps the scraped-data question alive even as the marketing shifts toward licensed partnerships. v6 is a signal of direction, not a closed case.