Right of publicity is a fifty-state mosaic
Right of publicity protects an individual's commercial interest in their name, image, voice, and other identifying attributes. It is creature of state law, and the law differs dramatically across jurisdictions. About twenty states recognize the right by statute; another fifteen recognize it under common law; the rest are ambiguous or silent. Statutory protections range from California's expansive coverage of name, voice, signature, photograph, and likeness (Cal. Civ. Code § 3344) to states with no statute and limited case law.
The patchwork creates predictable problems for synthetic-media products. A synthetic voice product distributed nationally may infringe in California, be unaddressed in Wyoming, and operate in legal gray space everywhere in between. Forum-shopping by plaintiffs is the standard response, but defendants can't reciprocally select where they distribute. The result is that California law often functions as a de facto floor.
Voice specifically is the most variable element. California, Tennessee, Indiana, and a handful of other states cover voice expressly. Most do not — voice claims in those states proceed by analogy to likeness, which works imperfectly because voice and visual likeness are distinct identifiable attributes.
II · Foundational casesThe cases that shaped voice as a protectable attribute
Two cases anchor the modern doctrine. Midler v. Ford Motor Co. (1988) held that imitation of Bette Midler's distinctive voice in an advertisement, by a singer hired specifically because she sounded like Midler, violated her common-law right of publicity. The Ninth Circuit reasoned that voice, like visual likeness, is a unique identifier whose unauthorized commercial use is actionable. Waits v. Frito-Lay, Inc. (1992) extended the analysis to imitation that did not literally copy a recording.
The doctrine is well-suited to AI voice cloning, which often produces something legally indistinguishable from Midler-style imitation: an audio output specifically engineered to sound like the plaintiff's voice, used in commerce, without consent. The novel issue is scale: AI voice cloning operates at zero marginal cost, so the question shifts from individual lawsuits to systemic enforcement.
The pending BBC v. Neural Labs case in the UK High Court is the first major case to test these doctrines in the AI-cloning context at scale. The complaint alleges unauthorized cloning of multiple BBC presenters for commercial use. The case will likely be settled or yield a ruling that maps neatly onto Midler's framework, but its significance is the practical demonstration that voice cloning can produce six-figure liability per voice.
III · FederalThe NO FAKES Act and the case for uniformity
The NO FAKES Act (Senate Bill 4875, 2024) would create a federal “digital replica right” — an exclusive right against unauthorized AI-generated reproductions of an individual's voice or visual likeness. The bill provides for statutory damages, injunctive relief, and a mechanism for individuals to assert the right against platforms and providers.
The bill's appeal is uniformity: a single federal standard would eliminate the forum-shopping problem and provide a nationwide remedy. Its risks are equally clear: First Amendment concerns about overbroad coverage of public figures, particularly in news reporting and parody contexts; difficulty defining “digital replica” with precision; and the question of how to handle de minimis or transformative uses.
Earlier drafts of the bill faced criticism for sweeping in legitimate satire and journalism. Successive revisions added news-reporting and First Amendment safe harbors, but the boundaries remain contested. The bill has bipartisan sponsorship and significant industry support (entertainment unions in particular) but has not yet passed either chamber. A scaled-down version focused narrowly on non-consensual sexual deepfakes — a less controversial subset — has moved further but does not address the commercial cloning problem.
IV · First AmendmentFree speech limits on right of publicity
Right of publicity is constrained by the First Amendment. Courts apply different tests in different circuits — the “transformative use” test (California), the “predominant use” test (Sixth Circuit), the Rogers test from trademark law (Second Circuit) — but the through-line is that newsworthy uses, commentary, satire, and substantial creative transformation receive protection that pure commercial appropriation does not.
For AI voice cloning, the doctrinal question is whether a synthetic version of a public figure's voice in a creative work — a podcast that uses synthetic voices to dramatize historical events, for instance — is transformative or appropriative. The analysis is highly fact-specific. A documentary that synthesizes a deceased historical figure's voice to read their own writings is likely protected; a commercial that uses a synthesized celebrity voice to endorse a product is almost certainly not.
The hardest cases sit between these poles. Synthetic voice cloning for legitimate journalism (e.g., dramatic reading of leaked communications) raises difficult First Amendment questions that haven't been definitively resolved. The pending federal legislation has tried to draft around these scenarios with explicit safe harbors; whether the safe harbors are wide enough is the most contentious drafting issue.
V · PracticalCompliance for synthetic-media products
For products that generate or distribute synthetic voice or likeness content, the practical compliance posture has four elements:
- Verified consent for any identifiable real person. Don't rely on user-supplied permissions; build verification into the workflow. Voice actors and licensed talent provide a clean source; scraped audio of public figures does not.
- Detection of unauthorized cloning attempts. Block prompts that name real people; implement voice-fingerprinting against a registry of known protected voices.
- Clear channels for takedown and rights assertion. Individuals must be able to assert that their voice has been cloned and have content removed promptly. The mechanism should be at least as user-friendly as DMCA takedown, ideally more so.
- Geographic awareness in distribution. Certain content that is legally permissible in one jurisdiction may be infringing in another. Geographic gating where appropriate, and clear contractual allocation of liability with downstream distributors.
The exposure for a product that ignores these is not theoretical. Statutory damages under California's right of publicity start at $750 per violation. Aggregated across millions of synthetic outputs, the math is unforgiving.
SD Frivolous Editorial
The SD Frivolous editorial team combines legal practitioners, journalists, and technologists focused on AI content law. Analysis is peer-reviewed by counsel before publication.
This analysis is journalism and commentary, not legal advice. Laws governing AI content change rapidly. Consult qualified counsel for specific legal questions.