Copyright of AI-Generated Music

A song normally carries two copyrights: one in the composition (melody, harmony, lyrics) and one in the sound recording. AI generation complicates both. In the United States, the Copyright Office requires human authorship, and courts agree: in Thaler v. Perlmutter, the D.C. Circuit in 2025 upheld the refusal to register a work described as created autonomously by an AI system. The Office's January 2025 report on copyrightability concluded that prompts alone generally do not give a user enough control over the output to be its author, while human-written lyrics, human-played parts, and creative selection, arrangement or modification of AI material can be protected. Applicants are expected to disclose AI-generated material that is more than minimal. In practice, a track produced from a text prompt in a service like Suno or Udio may contain little or nothing protectable. A song with your own lyrics over an AI backing track protects your lyrics and arrangement choices, not the generated audio. A common misconception is that a platform's terms of service decide copyright. Terms can grant you commercial use rights or promise the company will not claim ownership, but a contract cannot create a copyright the law does not recognize. Rules differ abroad: the UK has a provision for computer-generated works, which the government has been reviewing. Streaming is a separate layer. Distributors often accept AI tracks and royalties flow under distributor and platform agreements, but services prohibit spam uploads, artificial streaming and impersonation of real artists, and some label AI content. In June 2024, Universal, Sony and Warner sued Suno and Udio, alleging the services copied recorded music at scale to train their models; the companies argued training was fair use. In late 2025 some labels announced settlements and licensing arrangements with the services, while other claims continued.