Universal and Sony allege that Suno used 60,202 recordings to train its AI music models, including songs by BTS, Psy and 2NE1. The report describes evidence presented in the US copyright case, but the allegations have not been ruled on by a court.
Suno Watch analysis
What happened
The Korea JoongAng Daily says the list was submitted in September in litigation against Suno. It reports that experts hired by the labels examined Suno’s source code and training data, then compared audio fingerprints with recordings the companies control or represent in the US. The list includes Psy’s “Gangnam Style” and “Gentleman”, and BTS tracks including “Bad Decisions” and “Waste It On Me”.
Suno’s 20 October response, as described in the article, says the company obtained audio from YouTube for training and argues that its process produces a new, non-infringing product protected by fair use. The court has not decided the copyright claims. The report also says Warner Music Group settled its earlier case with Suno, while Universal and Sony pursued further litigation.
Why it matters
This is a concrete dispute over what music was used to train a commercial generative-AI system, and how evidence about that training is gathered. The labels’ list is not a court finding, and it covers recordings over which they say they hold US rights, not every track Suno may have used.
The report says Suno acknowledged using YouTube audio, while contesting the legal significance of its training process. That leaves the central question where it belongs: with the court, not the headline writers.
Our read
The reported examination of training data makes this a more specific account than a general accusation about AI and copyright. But a detailed allegation is still an allegation. The useful distinction is between what the filing reportedly identifies and what a judge has actually decided: at present, the latter is nothing.
What to watch
- Suno’s further filings and any response to the labels’ evidence.
- Whether the court accepts the labels’ account of how the recordings were identified.
- Any ruling on Suno’s fair-use argument, and whether Korean rights holders pursue related claims.
Discussion spark: If AI companies disclose that they trained on publicly available audio, is that enough, or should they have to identify the recordings and obtain permission?
Sources and evidence
- Source update (28 October 2026, 05:52 UTC)
not affiliated with or endorsed by Suno