New Paper Exposes Legal Gap in AI Music: Lyrics vs. Voice Cloning
Unauthorized lyrics copying is clear infringement, but AI voice imitation falls through federal cracks.
A new technical-legal paper by Zuhaib Hussain Butt, published on arXiv, systematically examines how generative AI music systems (e.g., Google Gemini's music tools) interact with U.S. copyright law under 17 U.S.C. Title 17. The analysis dissects specific technical components—prompt encoding, latent diffusion models, neural vocoders, and speaker embeddings—and maps each to legal risks. Central finding: unauthorized copying of lyrics presents a high infringement risk for musical compositions, whereas mere AI voice imitation typically falls outside federal sound recording protection (17 U.S.C. Section 114) and instead implicates state-level publicity rights, as seen in emerging laws like Tennessee's ELVIS Act. The paper cites recent cases (Concord v. Anthropic, Kadrey v. Meta, Lehrman v. Lovo, UMG v. Uncharted Labs) to illustrate this legal split.
The paper identifies a critical regulatory gap: federal law robustly protects lyrics and melody but offers limited remedies for synthesized vocal likeness. While state voice-cloning laws (like the ELVIS Act) are emerging, the author argues that a uniform federal framework is needed. The paper concludes with policy suggestions for clearer rules on AI music creation, urging lawmakers to update copyright doctrine for the age of synthetic media.
- Copying protected lyrics into AI music systems is high-risk infringement under 17 U.S.C. Section 106.
- AI voice imitation is not protected federally; it falls under state publicity rights like Tennessee's ELVIS Act.
- Technical components (latent diffusion, neural vocoders) are mapped to specific legal exposure points.
Why It Matters
Clarifies legal boundaries for AI music tools, impacting creators, platforms, and future federal copyright reform.