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

Sony Music and UMG sued AI music generator Suno over its v6 models, alleging they indirectly relied on outputs and user-preference data from earlier models trained on unlicensed recordings. The labels claim infringement involving at least 60,202 sound recordings, creating potential statutory damages of up to $9 billion, plus up to $2,500 per alleged circumvention of YouTube anti-downloading technology. Suno disputes the claims and says v6 was trained on licensed partner content, community interactions and internal learnings, but the case heightens legal risk for AI models using synthetic or derivative training data.

Analysis

The economically important issue is not statutory damages but whether a licensed-data launch can be contaminated by feedback loops tied to legacy, allegedly infringing models. If that theory survives an early motion, every AI-music vendor will need auditable dataset lineage, model-output provenance, and indemnification; this raises compliance costs and shifts bargaining power toward owners of large, clean catalogues. UMG.AS and SONY gain leverage in renewal negotiations, while WMG's early partnership may become less differentiated if rivals are forced into equivalent licenses.

Near term, the lawsuit is unlikely to materially move reported music-publishing revenue, but it strengthens the labels' ability to demand minimum guarantees, usage-based royalties, and audit rights over the next 1-3 months. The more consequential 6-18 month outcome is that generative tools may become a new high-margin licensing channel rather than a substitute for recorded music—provided courts do not validate a broad fair-use defense for training. SONY's diversified earnings base limits direct upside; UMG.AS offers cleaner exposure to a favorable licensing precedent, while WMG faces the most relative risk if its existing agreement proves to have conceded economics before industry standards are set.

The contrarian point is that a settlement could be more bullish for Suno's licensed partners than a decisive legal victory: labels may prefer recurring platform economics and attribution controls over an injunction that drives creation activity to offshore or open-source models. The thesis is falsified if discovery establishes genuinely segregated training data and no economically meaningful reuse of prior-model outputs, or if a court rejects output/interaction data as derivative infringement. Watch for disclosed AI-license minimum guarantees, royalty-rate language, and any change in label guidance on digital/ancillary revenue rather than headline damage estimates.

Key Decisions for Investors

• Prefer long UMG.AS versus short WMG over a 3-6 month horizon: UMG has greater leverage to reset industry-wide AI licensing terms, while WMG's incumbent deal risks becoming the benchmark before litigation clarifies required protections. Target 10-15% relative return; exit if WMG discloses economics materially superior to peers or UMG does not cite AI licensing as a revenue contributor by the next two reporting cycles.
• Maintain a tactical long SONY ADR (SONY) only as a lower-beta legal-optionality position, not a pure AI-music trade. Size modestly because music upside is diluted by games, imaging and financial exposures; reassess after the next earnings release for evidence that management is monetizing catalog rights rather than merely defending them.
• Avoid treating DEEZR as a direct beneficiary absent confirmation of its current public-market investability and contractual exposure to generative-music licensing. Use it as a watch item: stricter provenance requirements could raise moderation and content-verification costs for streaming platforms before they receive offsetting licensing revenue.
• Set an event alert for dismissal, preliminary-injunction, or discovery rulings over the next 1-3 months. A ruling that model feedback derived from legacy outputs can taint a licensed successor model would justify increasing the UMG.AS/WMG relative-value position; an early dismissal would remove the near-term catalyst and argues for closing it.