Explainer
A short history of AI music
Most people think AI music started the day a chat window wrote them a chorus. It actually started in a university lab in the 1950s, with a computer the size of a room writing for string quartet. What changed recently is who gets to use it and who gets sued over it.
Here's the timeline, with sources. The lawsuits get only a quick pass here; they're a story of their own.
The lab years
1956–57: the ILLIAC writes for strings. At the University of Illinois, Lejaren Hiller and Leonard Isaacson used the ILLIAC I computer to generate the Illiac Suite, using rules and probabilities to pick the notes. It's widely regarded as the first major score composed with a computer. The first three movements premiered on Aug 9, 1956, and the finished work is dated 1957 (Illinois Distributed Museum; Hiller & Isaacson, Journal of the Audio Engineering Society).
Note what the machine produced: a score, which people then had to play.
The model years
Apr 30, 2020: OpenAI's Jukebox. A model that generated raw audio, including rough singing, conditioned on genre, artist, and lyrics. OpenAI released the code, the weights, and samples (OpenAI; paper). It was slow and murky, but it was sound, not sheet music.
Jan 2023: Google's MusicLM. It generated music from a text description and could hold together for several minutes (paper). Google said at the time that it wouldn't release it (TechCrunch, Jan 27, 2023).
May 2023: platforms start writing rules. Audius added an opt-in "AI-Friendly" setting that let artists allow AI tracks trained on their music, filed under a "Generated with AI" section on their profiles (Music Ally, May 3, 2023). The same week, a soundalike track imitating two chart-topping singers was taken down from Audius within about an hour after the platform contacted the rights holders (Billboard).
The everybody years
Dec 2023: Suno goes public. Suno launched a public web app, and Microsoft added Suno-powered song creation to Copilot (TechCrunch, Dec 19, 2023). Now anyone could type a sentence and get back a full song with vocals.
Mar 21, 2024: Suno v3, released to all users, with full two-minute songs (Suno).
Jun 24, 2024: the labels sue. Universal, Sony, and Warner sued Suno in Massachusetts and Udio in New York, in cases announced by the RIAA, alleging mass copying of sound recordings to train their models (Reuters; Suno complaint).
The deal years
Oct 29–30, 2025: Universal and Udio settle, and agree to build a licensed AI music platform (UMG; Reuters). Udio has since signed deals with Warner, Merlin, Kobalt, Believe, and the NMPA, which leaves Sony as the only major without one (MBW, Sep 22, 2026).
Nov 2025: Warner settles with Suno and signs a licensing deal. Warner was dismissed from the case in December 2025 (MBW, same article).
Aug 12, 2026: BMG signs with Suno. Sep 8: Believe partners with Suno. Sep 9: Suno v6 launches (MBW, same article). We reviewed it here.
Sep 18, 2026: round two. Universal and Sony file a second suit against Suno, this one over 60,202 recordings and aimed at v6. In response, Suno says v6 was trained on licensed content, on "interactions including creations and preference signals" from its users, and on its team's "accumulated learnings" (MBW, same article). The original 2024 case is still going in Boston.
What the timeline says
Three things stand out.
- The technology is old. Access is new. It took about 65 years to get from the ILLIAC to a song generator anyone could use in a browser.
- The fight is about training data, not about machines making music. Every lawsuit above is about what the models learned from. Nobody sued the ILLIAC.
- The lawsuits are turning into licensing deals. In under a year, Suno signed with Warner, BMG, and Believe, and Udio settled with Universal. The labels now point to those deals in court as proof that a licensing market exists (MBW, same article).
For a sense of how far the tools have come in just the last year and a half, compare our Suno v6 review with the v4.5 era.
Sources checked Nov 3, 2026. Lawsuit details change quickly.