DRULES AI
New evidence confirms millions of copyrighted songs, from chart-toppers to deep catalog cuts, were systematically fed into Suno and Udio without licenses from most rights holders.
🗂️ Atlantic's Database Bombshell
Four searchable catalogs totaling roughly 21 million tracks have surfaced through investigative reporting, documenting exactly which recordings trained the leading AI music generators. The largest dataset holds 12 million songs, another nearly 9 million, with two smaller ones around 100,000 tracks each. Sony, Universal, and Warner are pursuing damages up to $150,000 per infringed composition and recording.
Suno previously admitted training on "tens of millions" of recordings. The fresh transparency has ignited a new wave of anger among professional creators, many of whom took to X yesterday to confront the company directly. One composer with significant engagement reminded Suno of its 2024 court filing admitting it scraped essentially all reasonably high-quality music files available online.
💼 Legal Chess Moves Continue
Warner Music Group settled its Suno case in late 2025, licensing its catalog for AI development while promising opt-ins for artists on voice, likeness, and composition rights. Universal's litigation against Suno remains active. The American Federation of Musicians recently sued both UMG and WMG, claiming the labels failed to share settlement proceeds or future licensing revenue with the musicians who actually performed the works now powering AI models.
AI companies maintain their use qualifies as fair use for transformative purposes. Labels counter that the systematic reproduction of entire catalogs constitutes piracy at commercial scale. A parallel book-publishing settlement reached $1.5 billion, offering a possible roadmap though music's mechanical and master rights add complexity.
🛡️ Artists Fight Back With Tech
Researchers at the University of Tennessee introduced HarmonyCloak, an audio tool that embeds imperceptible perturbations into recordings to poison AI training data without affecting listening quality. Independent artists are exploring similar self-defense measures while pushing for clearer legislation and platform-level detection of AI-generated content.
On the creation side, Suno users reported access to a v4.5-all free model over the weekend, with some noting automatic default artwork added to tracks. The juxtaposition is stark: hobbyists and indie creators enthusiastically experiment with the latest version while professionals whose catalogs built the underlying models demand accountability and compensation.
The conversation on X mixed personal releases, calls for regulation similar to sample clearance, and skepticism that any lawsuit money would actually reach working artists rather than private equity owners of the labels.
Bottom line: Concrete proof of 21 million scraped tracks has crystallized the fight, forcing the industry to decide whether AI music will be built on consent and revenue sharing or continue exploiting existing catalogs.