What happened
Suno filed its 44-page answer on October 20, 2026, in the U.S. District Court for the District of Massachusetts. The case, *UMG Recordings, Inc. v. Suno, Inc.*, is brought by UMG, Capitol Records, Sony Music Entertainment and affiliated labels. The complaint was filed on September 18, 2026. The answer responds paragraph by paragraph; it is Suno’s position in litigation, not a court finding.
The filing admits that Suno’s generative service lets users create digital music files from basic inputs and that its model was built by showing a program many kinds of sound recordings to derive statistical insights. Most directly, Suno admits it obtained audio data from YouTube for use as training data. It also says it has not publicly disclosed the contents of its training data, which it regards as highly confidential. The answer describes the company’s v6 model as trained on content licensed from partners, community interactions—including creations and preference signals—and accumulated learnings from its team.
Suno denies allegations that it set out to supplant human creativity or substitute existing works with generated substitutes. It says that, since January 2026, users may include artist names in prompts, but claims those names are converted into style descriptions before generation. The filing quotes Suno’s explanation that it removes an artist’s name and redirects the request toward descriptive musical characteristics, and says the company chose not to teach models artist names. These are the company’s representations in its answer; the filing does not independently establish how the system behaves.
The answer also acknowledges Suno’s partnerships with Warner Music Group, BMG and Believe, including development of models using recordings licensed from those partners. It states that Suno raised more than $400 million in a 2026 Series D at a $5.4 billion post-money valuation, and offers a Premier subscription for $24 per month. Suno denies or says it lacks sufficient knowledge to admit many of the plaintiffs’ factual allegations, and asserts fair use as a defense. A defense is a legal argument, not a ruling on the merits.
What the archive adds
New Runtime’s earlier coverage of Suno’s answer in the earlier lawsuit noted that Suno admitted using YouTube audio as training data while denying other allegations. This filing is a new event in a later case, but repeats that central admission and adds a new set of plaintiffs and claims.
Trend signal
This does not yet fit a pattern New Runtime tracks. The filing is a concrete legal development about one company’s training-data practices, not evidence of a broader operational pattern across AI systems.
Why now
The answer responds to a complaint filed on September 18, 2026, and was entered on October 20. It gives Suno’s formal responses and defenses as the new case proceeds, while leaving the underlying copyright dispute for the court to resolve.
Connections to other events
The filing itself connects the current dispute to the earlier UMG case against Suno, filed in June 2024, and describes prior litigation over training data. Suno also points to licensing partnerships with Warner Music Group, BMG and Believe as part of its account of model development. Those partnerships do not resolve the claims about the training data at issue here.
Why it matters
For teams building generative products, data provenance is not an abstract compliance concern: it can become a central factual question in litigation, alongside how a model was trained and what records the company can produce. Suno’s admission about YouTube audio is significant because it identifies a source category, while its confidentiality position and denials show that a public statement about training data may remain incomplete or contested.
The filing is not a legal precedent and does not establish that training on the audio was unlawful or permissible. But it is a practical reminder to preserve dataset inventories, acquisition records, licenses, and model-specific lineage. If a product depends on third-party media, teams should be able to distinguish licensed material, user-provided signals, and other data sources—and have counsel assess the rights and documentation for each.
Limits and uncertainty
The answer records Suno’s admissions, denials, and defenses, not findings by the court. The filing does not settle whether the YouTube audio was copyrighted, whether particular recordings were included, or whether the training use qualifies as fair use. Suno says its training-data contents are highly confidential, and the plaintiffs’ claims remain disputed. The court has not determined liability or ordered a remedy in this answer.
Reader takeaway
Suno’s answer confirms that YouTube audio was among its training inputs, but the admission does not decide the copyright question. For builders, the durable lesson is to maintain auditable data lineage and licensing records before a dispute makes those records central.
Open question
How will the court evaluate Suno’s fair-use defense alongside the admitted use of YouTube audio for training?