Source: Vanderbilt Law School

Paul Lehrman heard in 2023 a synthetic rendition of his voice reading words he never spoke, marketed under the alias “Kyle Snow,” which the piece attributes to an AI model by Lovo trained on recordings Lehrman had previously submitted to a person who was actually a Lovo employee. Lehrman and voice actor Linnea Sage sued, and the Southern District of New York’s 2025 proceedings (Lehrman v. Lovo) provided an early judicial test of existing law’s treatment of AI voice cloning. The court concluded that federal intellectual property law provided little protection: copyright protects fixed recordings rather than an unfixed human voice, and an AI model that mimics pitch, cadence, and timbre without reproducing a particular recording does not infringe; similarly, the Lanham Act claim failed because the voices were treated as the actors’ product rather than source-identifying trademarks.

The opinion relied on Midler v. Ford Motor Co. as doctrinal backdrop but noted Midler involved a human sound-alike rather than a model trained on an actor’s own recordings, leaving unsettled whether a clone treated as a derivative work might lead to preemption of state claims. The court allowed a New York Civil Rights Law claim to survive in a footnote, holding it was not preempted by the Copyright Act because it is grounded in identity rather than a fixed work. State-law variation is decisive: New York’s statute covers only the direct user for advertising or trade, while Tennessee’s ELVIS Act extends liability to distributors of voice-cloning technology. Contractual protections exist—SAG-AFTRA’s 2025 Interactive Media Agreement imposes consent-and-compensation requirements for covered performers—but that protection does not reach most voice actors. A 2026 National Association of Voice Actors survey reported 21% of respondents said they lost work directly to AI (up from 14% the prior year). The article notes the NO FAKES Act pending in Congress could create a federal right for unauthorized digital replicas of voice and likeness, potentially preempting or supplementing the current state-by-state patchwork, but leaves unresolved whether Congress will act and how courts might treat model-trained clones in preemption analyses. The piece frames its conclusions with precise attribution to the court rulings, statutes, union agreements, and surveys cited in the source.