Topic
Two US judges ruled in June 2025 that training language models on books was fair use, but only on the records in front of them, and one said plainly that most such copying will likely be illegal. Music cases turn on market harm, and that is where songwriters and labels have a stronger hand than the authors did.
Artist rightsvia US District Court, N.D. California (via CourtListener RECAP)
In the US, a prompt on its own does not make you the author of what an AI generates. Lyrics, melodies, performances, edits and creative arrangement you made yourself can still be registered, as long as you claim your part and disclose the AI part on the form.
Artist rightsvia U.S. Copyright Office
In April, Believe called Suno a "pirate studio" and said its detector could spot Suno tracks with 99% reliability. On 8 Sep 2026 it signed a global partnership to distribute music from Suno's new licensed model. The output didn't change category because of the technology. It changed because of a contract artists weren't party to.
Industryvia Suno
The US Justice Department has told the court in the New York Times' case against OpenAI that training AI models on copyrighted text is fair use, and it goes out of its way to reject the "market dilution" theory the music industry leans on. It leaves outputs, data sourcing and identity open, and that is where artists' claims are strongest.
Artist rightsvia US Department of Justice (copy hosted by MusicTechPolicy)
Moises has turned its stem-separation app into a browser workspace with multitrack recording, MIDI, generated parts, private voice models and an agent, and is plugging the same tools into Fender Studio Pro, UA LUNA and SoundCloud. The training promise is unusually firm; the terms on outputs and storage deserve a closer read.
Releasesvia Moises Newsroom
One email a week: the music-tech and AI-in-music news that mattered, and what we think about it.