Fetching from the wire…
Policy2026-09-21 · source-backed
UMG and Sony filed in the District of Massachusetts after a judge refused to let them amend the original June 2024 suit, naming 60,202 specific recordings surfaced by forensic analysis of training data. The novel argument is that Suno's licensed v6 models are tainted because they were trained partly on user outputs from the earlier unlicensed models. At the $150,000 willful-infringement ceiling the named works carry a theoretical maximum above $9B. The distillation-taint theory threatens any model whose clean successor was trained on the dirty one's outputs, which is a lot of models.
Each link below shares sources, entities, or timing with this story.
A coalition including Universal, Warner, Sony, HYBE, Believe, BMG, Concord, Mom+Pop, Partisan, Dirty Hit and Glassnote proposed four eligibility principles: the AI platform must be authorized, the track must comply with copyright and personality rights, AI use must be labeled,...
CEO Mikey Shulman announced audio watermarking and fingerprinting plus a download policy limiting mass distribution to streaming platforms, targeting schemes where users generate thousands of tracks, upload them to DSPs and bot the plays for fraudulent royalties. Suno also sig...
V6, V6-wild and V6-mini launched September 9, about ten months after the first WMG deal, with revenue sharing starting on launch day and distribution to individual artists left entirely to the parent organizations. Suno declined to disclose dataset size, artist representation...
The Verge (corroborated by TechCrunch) reports a hacker used an employee's credentials to access Suno's source code, exposing pipelines that scraped millions of songs and lyrics from YouTube Music, Deezer, and Genius. The discovery mechanism is the story. Discovery in the labe...
Music generation splits into a symbolic planning step writing an editable melody and chord score, then audio synthesis rendering vocals and accompaniment, so a user revises the composition before hearing it. YuE2-3B is about 3.59 billion parameters over 28 layers, with MERT2 e...
North Small Translate is a mixture-of-experts model with 218B total and 25B active parameters, covering 50 languages. It scores 83.60 on WMT26, and an agentic multi-pass variant scores 84.36. DeepL NextGen scores 81.37 and Google Translate 68.20. On book-length translation in...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.