Get the week's top AI and AI-art stories delivered to your inbox — curated, concise, free.
Free. Unsubscribe any time.
Sony Music Entertainment has sued AI music generator Udio in a New York federal court, alleging copyright infringement across more than 30,000 songs — a catalog stretching from Elvis Presley's "Hound Dog" to Beyoncé's "Say My Name" to Harry Styles' "As It Was."
30,000 songs is not a symbolic number. It represents a substantial slice of Sony's active commercial catalog, and filing a lawsuit that names specific tracks — rather than speaking in generalities about "training data" — is a deliberate legal strategy. Specific song titles make infringement easier to demonstrate and damages easier to quantify. Each registered copyright can carry statutory damages of up to $150,000 for willful infringement under U.S. law, which means the theoretical exposure here runs into the billions.
According to The Verge, the suit was filed Monday in a New York court and alleges that Udio's AI music generator was trained on Sony's copyrighted recordings without authorization.
This is not the first time major labels have gone after AI music generators. A coalition of labels — including Sony, Universal, and Warner — filed suits against both Suno and Udio in mid-2024. What's notable here is that Sony is now filing an additional, standalone action specifically against Udio, with a named catalog of this scale. That suggests the earlier litigation either didn't resolve the underlying training-data question to Sony's satisfaction, or that Sony is building a parallel record for a separate damages calculation.
The timing also matters. Suno's own legal exposure deepened after hacked source code revealed the company had scraped millions of songs from YouTube Music, Deezer, and Genius — a disclosure covered in detail in our earlier report on Suno's training data scraping practices. That story made clear that AI music generators have largely avoided transparency about their training sets, and courts are now forcing the issue.
For creators building music into AI-generated video, game audio, or companion experiences, the legal environment around AI-generated audio is tightening in a specific way: it's not just about the output sounding like a known artist, it's about whether the model was trained on protected recordings in the first place. Platforms that can't demonstrate clean or licensed training data are increasingly litigation targets.
That creates a real workflow consideration. Tools that have disclosed licensed training sets — or that generate audio from synthesis rather than interpolation of existing recordings — carry meaningfully lower legal risk for the platforms hosting them, and arguably for the creators publishing content made with them. As these suits proceed, expect AI music platforms to either publish clearer data provenance statements or face pressure from investors and enterprise customers to do so.
Udio has not yet issued a public response to this specific filing. The case will likely take years to resolve, but each new lawsuit adds to the legal record that courts and Congress will use to eventually define what "fair use" means for generative AI training — a question that affects every modality, not just music.
For now, creators building audio-driven projects should watch which music-generation platforms are named in active litigation and factor that into tool selection, particularly for commercial work.
