Suno has filed a motion to impound the precise number of audio files used to train its generative AI models, escalating its bitter copyright battle with UMG and Sony Music. The request, submitted May 29 2026 in Massachusetts federal court, targets only the aggregate "Model Training Figure" rather than specific tracks.
⚖️ The Courtroom Power Play
Documents reveal Suno arguing that revealing the dataset volume would let competitors benchmark performance, infer architectural choices, and gain unfair advantages in the fast-moving AI music space. The company claims the figure qualifies for impoundment under prior court orders requiring good-cause showings, which it says it has already met. UMG and Sony are not currently opposing the seal but reserve the right to contest it later. Separately, the labels want to add 61,026 additional copyrighted works to their infringement claims. Suno intends to oppose that amendment.
📊 Training Data at the Center of Everything
The dispute shines a harsh light on how these models are built. Reports tied to the case suggest Suno ingested millions of tracks scraped from across the open internet. Related filings have surfaced claims that the company essentially trained on "all available music files" online, reigniting fury among rights holders. Warner Music settled with Suno in November 2025, but licensing negotiations with UMG and Sony have stalled with "no path forward" according to sources. The court has approved similar impoundments in other AI cases, and dispositive motions aren't due until January 2027.
🛠️ What Creators Should Watch
For producers and artists using Suno daily, this isn't distant legal theater. Platform stability, future feature access, and pricing all hang on these outcomes. A victory for Suno on training data secrecy could embolden other generative tools while delaying transparency around fair use. Yet if the labels force disclosure and win broader infringement rulings, we could see stricter content filters, higher subscription costs, or limited export rights. Smart creators are already stress-testing workflows across Suno, Udio, and emerging alternatives while building audiences that value their curation skills over raw generation. The flood of AI output is real. Deezer alone sees roughly 60,000 AI-generated tracks daily, many flagged and removed, showing the supply-side pressure is only increasing.
The case also highlights shifting industry power dynamics. After suing, some majors pivoted toward potential licensing deals, treating AI platforms as both threat and inevitable distribution partner. Suno's aggressive posture suggests it believes its technology edge is worth protecting at all costs, even if it means fighting in court for another year.
Bottom line: Suno's bid to keep its training dataset size secret reveals how data scale is now the most valuable and most dangerous asset in AI music development.
DRULES AI