AI Voice Cloning Gap Leaves U.S. Performers Legally Exposed

Artificial intelligence tools capable of replicating a performer's voice with near-perfect accuracy are now available to commercial buyers for as little as a few hundred dollars per month, yet no federal statute in the United States specifically prohibits their use without a performer's consent. That legal vacuum is drawing urgent warnings from entertainment unions, legal scholars, and digital rights advocates who say the gap leaves working actors, voice artists, and broadcasters exposed to exploitation on an industrial scale.

The Technology and How It Works

Voice cloning, at its technical core, involves feeding a machine-learning system — typically a neural network trained on large volumes of audio — enough recorded samples of a specific person's voice to allow the system to generate new speech in that person's vocal style. The process, known as text-to-speech synthesis with speaker adaptation, has advanced rapidly. Where earlier systems required hours of training audio, current commercial platforms can generate a convincing synthetic voice from as few as three to thirty seconds of source material.

From Novelty to Commercial Infrastructure

Until recently, high-fidelity voice cloning was confined to well-resourced film studios and specialist post-production houses. That constraint has dissolved. Platforms marketed to advertising agencies, podcast producers, audiobook publishers, e-learning developers, and corporate communications teams now offer voice cloning as a standard feature. According to reporting by Wired, several of these platforms explicitly advertise the ability to create a "digital voice twin" without requiring ongoing participation from the original speaker — a selling point that cuts directly against a performer's economic interest.

The downstream implications are not theoretical. Voice actors who once earned residual income from repeat commercial use of their recorded performances can be replaced, at marginal cost, by a synthetic version of themselves licensed once and deployed indefinitely. For more on how similar technology intersects with hardware products, see how Meta limits synthetic voice features on consumer devices, a related debate over where platform boundaries should sit.

The Federal Legislative Gap

The United States has no omnibus federal law governing the right of publicity — the legal principle that an individual controls commercial use of their name, image, and likeness. Existing protections derive from a combination of state statutes, common-law doctrines, and contractual arrangements. The Lanham Act, which governs trademark and unfair competition, has been used in limited cases to challenge voice imitation, but courts have applied it inconsistently and its reach does not extend to non-commercial or transformative uses.

State-by-State Patchwork

California and New York offer the most developed right-of-publicity frameworks, but even these predate the mass commercialisation of generative AI and do not specifically address synthetic audio reproduction. Tennessee passed the Ensuring Likeness Voice and Image Security Act — known as the ELVIS Act — which does target AI voice cloning explicitly, becoming one of the few states to do so. Legal scholars writing in MIT Technology Review have noted, however, that state-level remedies are geographically bounded: a company operating from a jurisdiction without such a law, or contracting under the laws of a permissive state, may face no meaningful legal exposure regardless of where the affected performer lives or works.

Several federal proposals have been introduced in Congress, including legislation that would establish a baseline national right of publicity covering voice and likeness, but none has cleared committee as of this reporting. Industry lobbyists representing technology companies have argued that overly broad legislation could chill legitimate uses of AI in accessibility tools, language learning software, and documentary production, according to public testimony reviewed by this newsroom.

What Existing Contract Law Can and Cannot Do

For performers working under union agreements — primarily those covered by SAG-AFTRA — collective bargaining contracts have begun to include language addressing AI-generated performances. Those provisions, however, cover only signatories to those agreements. The vast majority of voice work in the United States is non-union: corporate narration, e-learning modules, regional advertising, and the burgeoning podcast and audiobook economy. Workers in those sectors have no collective agreement to fall back on and must negotiate individual contracts, most of which were written before synthetic voice reproduction was a practical commercial option. For background on how unions are approaching this evolving challenge, the debate around Hollywood unions and AI voice governance offers important context on the bargaining dynamics at play.

Industry Practices and Platform Accountability

The commercial voice cloning sector operates with limited standardisation around consent verification. Some platforms require users to upload a signed consent declaration before generating a cloned voice; others rely on terms-of-service provisions that place legal liability entirely on the end user. Independent audits of these consent mechanisms are rare, and enforcement against misuse is almost entirely reactive — meaning harm must occur and be reported before any remedial action is taken.

Consent Verification in Practice

According to IDC analysis cited in industry briefings, a significant proportion of voice cloning misuse involves internal corporate deployments rather than high-profile celebrity impersonation. A company may clone the voice of an in-house narrator or a contracted voice artist for ongoing use across internal communications, training videos, or customer-facing automated systems — without notifying or compensating the original performer. Because these deployments are not public-facing, they rarely surface in legal challenges or press coverage, leaving affected workers unaware their voice is in active commercial use.

Note: Platform features and policies are subject to change. Data compiled from publicly available product documentation and industry reporting. This table does not constitute an endorsement of any product.

The Human Cost to Working Performers

The performers most acutely affected by the legislative gap are not, in the main, the established stars whose names generate headlines. They are working professionals — voice actors who narrate audiobooks, record corporate training content, provide voices for video games, or staff telephone interactive-voice-response systems — for whom voice work constitutes a primary or significant supplementary income stream.

Gartner has projected that AI-generated voice content will displace a measurable share of traditional voice-over work within the current market cycle, with the steepest impact falling on mid-tier and entry-level performers whose work is most easily replicated by synthetic systems. The economic logic is straightforward: once a synthetic voice is created, its marginal cost of deployment approaches zero. A human performer, by contrast, charges per project, per session, or under residual agreements. The structural incentive for buyers is clear, and without legal constraint, market forces alone are unlikely to preserve performer livelihoods.

This displacement risk does not exist in isolation from broader technology-sector workforce questions. The disruption AI is causing to voice performers mirrors patterns seen across other knowledge-work categories — a dynamic examined in the context of large-scale AI-driven workforce restructuring at major technology firms, where automation decisions have similarly outpaced policy responses.

The Postmortem Problem: Voices Beyond Death

The legal exposure extends beyond living performers. The right of publicity in most U.S. states terminates at death or survives only for a defined statutory period — in California, seventy years; in other states, far less or not at all. AI voice cloning tools do not require a living subject. Archive recordings of deceased performers — available through broadcast archives, commercial releases, and digitised historical collections — are sufficient training material for generating new synthetic performances attributed to those voices.

This has already produced documented commercial disputes, with estates of deceased performers challenging the use of archived recordings to train or seed AI systems. The legal questions involved — including whether training a model on copyrighted recordings constitutes infringement, and whether the resulting synthetic output infringes a posthumous right of publicity — remain unresolved in federal courts. The broader legal terrain around deceased performers and synthetic audio is explored in detail in reporting on how AI voice cloning of deceased performers is testing U.S. digital rights law, a case study in how courts are beginning to grapple with questions that legislation has yet to answer.

What Reform Could Look Like

Legal scholars and policy advocates have proposed several frameworks for addressing the legislative gap. A federal right of publicity statute with explicit provisions for synthetic voice reproduction would establish a national floor, removing the jurisdictional arbitrage that currently benefits platform operators. Mandatory opt-in consent — requiring affirmative, documented agreement from a performer before their voice can be cloned for any commercial purpose — is among the most commonly cited structural requirements.

Some proposals go further, calling for a registration and licensing regime modelled on existing music rights frameworks, under which synthetic voice deployments would require a licence, generate a traceable record, and trigger royalty obligations. Critics argue such a system would impose compliance burdens on small developers and legitimate accessibility applications. Proponents counter that similar objections were made — and ultimately overcome — when digital music licensing was standardised.

The pace at which commercial AI voice tools are entering the market makes the timeline for reform consequential. According to Gartner and IDC projections, the volume of synthetic voice content in commercial circulation is growing at a rate that will make retroactive consent enforcement increasingly difficult. Each month without a federal statute is, in the assessment of digital rights advocates, a month in which the de facto standard becomes one of permissiveness by default.

Infrastructure policy, too, shapes who is exposed and who has recourse. The uneven distribution of digital connectivity — and the legal literacy that accompanies it — means that performers in rural and underserved markets may be disproportionately unaware of their exposure or unable to access legal remedies. Work to address foundational digital access gaps, such as initiatives examined in coverage of rural broadband expansion efforts in the American interior, has a longer-term bearing on whether all workers can meaningfully participate in whatever rights framework eventually emerges.

Until Congress acts, performers navigating this landscape will continue to depend on a combination of contractual vigilance, union membership where available, and the variable protections of their home state's law — a patchwork that legal experts, industry analysts, and labour advocates broadly describe as inadequate to the scale and speed of the technology it is being asked to govern. The market, in the absence of law, will continue to move in the direction of least resistance.