Is AI Voice Cloning Legal? Consent, Copyright & U.S. Law Explained

AI voice cloning can be legal in the United States, but consent is not the only legal test — and having consent does not mean everything else is automatically permitted. You must also have authority to use any source recordings, avoid deceptive or harmful uses, and follow the specific rules that apply to publicity rights, copyright, advertising, robocalls, contracts, and privacy. Some news, commentary, parody, satire, or other protected uses may be lawful without consent, but those exceptions are highly fact-specific. A voice itself is generally not protected by federal copyright, while a particular recording, script, or song may be. This article explains general U.S. principles as of July 14, 2026 and is not legal advice. Consult an attorney for guidance specific to your situation.

Key Takeaways

• Consent matters, but it is not the only test. Even with clear written consent, you must separately check copyright, advertising rules, robocall law, platform policies, and contracts.
• A person’s voice is not a copyrightable work under U.S. federal law. What may be protected is a particular sound recording, the underlying script or song, or creative human contributions embedded in an output.
• Voice identity is protected by state publicity and privacy law — and by the Lanham Act for false endorsements — not primarily by copyright. Midler v. Ford Motor Co. (9th Cir. 1988) and Waits v. Frito-Lay (9th Cir. 1992) both involved recognizable voices in commercial advertising, not all imitation generally.
• The NO FAKES Act (S.4591 / H.R.8915) was ordered reported favorably by the Senate Judiciary Committee on June 18, 2026. It has not yet become law. All rights and protections described in the bill are proposed, not current.
• Purely AI-determined output cannot be registered for copyright, per the Copyright Office’s January 2025 guidance — but scripts, music, human editing, and other contributions in the same work may still be protected.
• AI-generated voice robocalls require prior express consent under FCC rules interpreting the TCPA, regardless of whether any other law is satisfied.

What Laws Apply to AI Voice Cloning?

Voice cloning touches several distinct bodies of law. No single statute governs all uses.

Legal Layer	What It Protects	When It Applies	What to Check
State publicity / personality rights	A person’s voice identity, name, likeness	Commercial use of a recognizable voice without authorization	Authorization from voice owner; check your state and distribution state
Federal copyright	Sound recordings, scripts, songs, human-authored expression	Reproducing or distributing a protected work or substantially similar output	Who owns the source recording; whether the output is substantially similar
Lanham Act (15 U.S.C. § 1125(a))	Consumers and rights holders from false endorsement or passing-off	Any voice use implying a person’s sponsorship or approval	Is the use likely to cause consumer confusion about endorsement?
FTC Act (Section 5)	Consumers from unfair or deceptive acts	Deceptive ads, fake reviews, impersonation, false statements	Is the voice use deceptive or likely to mislead consumers?
FTC Impersonation Rule (2024)	Consumers from impersonating government entities or businesses	Voice clone impersonates a government agency or a real company	Does the voice clone falsely represent an official body or brand?
FCC / TCPA	Consumers from unwanted calls and AI-generated voices	AI-generated voice used in a telephone call to consumers	Prior express written consent required for AI voice calls under FCC’s 2024 ruling
Contract / platform terms	Platform and counterparty	Voice generated or distributed via a third-party service	Review the tool’s terms of service and any existing talent agreements
State privacy / anti-CSAM / election law	Additional state-specific interests	Varies by state and content type	Check laws in the state of creation and of the audience

Does Consent Make Voice Cloning Legal?

Consent is often central, but it is neither the only legal test nor a universal requirement. Some protected expression — such as news reporting, commentary, parody, or satire — may be lawful without the subject’s consent, depending on the specific facts. Conversely, consent from the voice owner does not automatically resolve copyright in a source recording, the rights in a script or song the cloned voice performs, applicable advertising rules, robocall regulations, or platform terms.

Who Can Give Consent, and for What?

Voice Situation	Consent Status	Key Risk Areas
Cloning your own voice for your own project	You consent to yourself	Still check: scripts, songs, platform terms, deceptive use
Colleague’s voice with clear written permission	Consent obtained	Ensure scope covers platform, duration, commercial use, modifications
Licensed voice talent via agent or union	Consent obtained with documentation	Confirm scope and that no additional rights are needed
Celebrity voice without permission	No consent	Publicity rights, Lanham Act, possible copyright in source recording
Deceased person’s voice	No consent (unless estate authorizes)	Post-mortem publicity rights vary by state; estate may control source recordings
Minor’s voice	Requires additional safeguards	Agreements involving minors may need parental/guardian consent, court approval, or compliance with child-performer laws, depending on state
Private individual without permission	No consent	Publicity rights, privacy law, potential harassment or fraud liability

What Written Consent Should Cover

A verbal agreement is difficult to prove and scope. Written consent is strongly recommended and should address:

• Whose voice is being cloned (full legal name)
• Which project(s) the voice will appear in
• Which platforms and territories the output may be distributed on
• Duration of the license (expiration or perpetuity)
• Commercial use — whether the output may be monetized
• Right to modify or edit the cloned output
• Sub-licensing — whether third parties may use the output
• Revocation procedure — how and when consent may be withdrawn
• Record-keeping — how and where the agreement will be stored

Consent does not solve everything. Even with a signed consent form, you must separately address: (a) copyright in source recordings or underlying works, (b) FTC/advertising rules if the output implies endorsement, (c) FCC/TCPA requirements for any telephone delivery, (d) contract terms of any platform or tool used, and (e) platform content policies.

Voice Rights vs. Copyright: What Is Actually Protected?

People frequently conflate four distinct categories. Treating them as one leads to either over-caution or under-caution.

A person’s voice is not itself a copyrightable work under federal law. What may be separately protected:

1. A sound recording (the specific fixed audio): owned by whoever produced it — often a record label or production company, not the performer. Using an unlicensed recording as source material may implicate reproduction rights, licensing terms, or other rights; whether a resulting output infringes depends on what protected expression was actually copied into the output.

2. The underlying composition — the script, lyrics, or musical work. Cloning a voice to perform a copyrighted song or read a copyrighted script requires a separate analysis of whether that underlying work is licensed, in the public domain, or covered by fair use.

3. Sufficiently human-authored contributions in the output — if a person made expressive creative choices (selecting, arranging, editing, scripting), those elements may be protectable even if the voice synthesis itself was machine-driven.

4. Purely AI-determined output: The U.S. Copyright Office reaffirmed in its January 2025 AI report (Part 2) that material generated without sufficient human authorship cannot be registered. This does not mean “anyone can freely copy the audio” — other rights (trademark, contract, platform terms, misappropriation) may still apply.

Current U.S. Voice-Cloning Laws in 2026

Tennessee’s ELVIS Act (Effective July 1, 2024)

Tennessee’s ELVIS Act (Ensuring Likeness, Voice, and Image Security Act, Public Chapter 588) was signed March 21, 2024 and took effect July 1, 2024. It amends Tennessee’s Personal Rights Protection Act to explicitly list “voice” as a protected attribute alongside name and likeness, and specifically addresses AI-generated replicas. Tennessee describes it as first-of-its-kind legislation targeting AI voice cloning. It applies to unauthorized use of AI to produce a voice replica for distribution.

California, New York, and Other State Laws

Several other states have relevant protections, though scope, duration, and exceptions differ:

• California: Existing Civil Code § 3344 (commercial use of name, voice, likeness) and post-mortem rights under § 3344.1, plus newer digital-replica bills targeting AI-generated replicas of deceased and living persons in specific contexts.
• New York: Civil Rights Law §§ 50–51 (commercial use) plus 2023–2024 legislation addressing digital replicas in entertainment and adult content.
• Illinois: Right of Publicity Act covers voice and likeness for living individuals.
• Indiana: Post-mortem right of publicity under Indiana Code § 32-36-1 covers voice; consult the current statute for the applicable duration and exceptions.
• Texas: Statute covering AI-generated likeness in specific deceptive contexts.

Rules are not uniform. The applicable law depends on where you are, where the content is distributed, and the nature of the use. High-stakes projects should verify each relevant jurisdiction.

The NO FAKES Act Is Still a Bill

The NO FAKES Act (“Nurture Originals, Foster Art, and Keep Entertainment Safe Act of 2026,” S.4591 / H.R.8915, 119th Congress) was introduced in 2026 and ordered reported favorably by the Senate Judiciary Committee on June 18, 2026, by voice vote. The White House’s March 2026 National AI Legislative Framework recommends a federal approach broadly similar in purpose, but that framework does not constitute formal endorsement of every provision of S.4591.

As of July 14, 2026, this bill has not been enacted. All descriptions below describe what the bill would do if enacted:

Feature	Proposed (Not Yet Law)
Who is covered	All U.S. individuals, not only celebrities
Protected attributes	AI-generated voice and visual replicas
First Amendment exceptions	News, commentary, parody, satire — with specific conditions; these exceptions are not blanket
Platform process	Notice-and-counter-notice; limitations on provider liability (not identical to DMCA safe harbor)

Are AI Voice Robocalls and Fake Endorsements Legal?

AI Voice Robocalls: FCC and TCPA

The FCC ruled in February 2024 that AI-generated voices in telephone calls constitute “artificial or prerecorded voice” messages under the TCPA (47 U.S.C. § 227). This means AI-generated voice calls to consumers require the same prior express written consent already required for robocalls generally. The consent type depends on whether the call is marketing (prior express written consent required) or informational (prior express consent may suffice). State robocall laws may impose additional requirements. Consult the FCC’s 2024 declaratory ruling for details.

Using a cloned voice to call consumers — even with a consent to use someone’s voice — does not satisfy TCPA consent requirements, which must come from the recipient of the call.

False Endorsement and FTC Rules

The FTC Act Section 5 prohibits unfair or deceptive acts in commerce. A voice clone that falsely implies a celebrity or public figure endorses a product or service may violate Section 5 and Lanham Act § 1125(a). The FTC’s 2024 Impersonation Rule specifically prohibits voice clones that impersonate government entities or real businesses.

Separately, state attorneys general have brought enforcement actions involving AI-cloned voices used in misleading robocalls and fake political endorsements. Each case has its own facts, dates, and statutes; no general claim that “several AG offices have prosecuted” should be taken as describing your jurisdiction without verification.

Voice-Cloning Risk by Use Case

The following table shows typical risk level and key things to verify — not a legal determination of lawful or unlawful, which depends on jurisdiction-specific facts.

Intended Use	Typical Risk Level	What to Verify
Cloning your own voice for your own project	Low	Source recording ownership; scripts/songs; no deception
Consented colleague’s voice for internal training	Low	Written consent scope; commercial vs. internal use
Parody of a public figure using a recognizable voice	Medium	Is it clearly satirical? Would a consumer think it is real? Is it used commercially?
Celebrity voice singing a new song for public release	High	Publicity rights authorization; sound recording license; underlying song license
AI voice used in a telephone marketing call	High	TCPA/FCC prior express written consent from call recipient
Cloning a deceased performer’s voice for a new release	High	Post-mortem publicity rights in applicable state(s); estate authorization; source recording license
Fake endorsement ad using a recognizable voice	High	Lanham Act § 1125(a); FTC Section 5; state AG enforcement risk
AI voice impersonating a government agency or company	Very High	FTC Impersonation Rule 2024; state law; criminal fraud statutes

High-risk projects — those involving celebrities, deceased persons, minors, political content, advertising, robocalls, or large-scale commercial distribution — should receive legal review before publication.

Consent and Publishing Checklist

Before publishing or distributing any content featuring a cloned voice:

• Do I have written consent from every person whose voice is cloned?
• Does consent explicitly cover the exact use, platform(s), territory, and duration?
• Does consent address commercial use, right to modify, and sub-licensing if applicable?
• Have I documented how and when consent may be revoked, and where the signed agreement is stored?
• Am I using source recordings I own or have licensed, not unlicensed studio tracks or third-party content?
• If the cloned voice performs a script or song, is that underlying work licensed, public domain, or covered by fair use?
• Have I verified that the output does not create a false impression of endorsement (Lanham Act / FTC)?
• If the content is delivered by telephone, have I confirmed TCPA/FCC compliance and recipient consent?
• Have I reviewed the platform’s terms of service for the tool used to create and for the platform where it will be published?
• Have I checked the state publicity laws of both the originating jurisdiction and the primary distribution jurisdictions?
• Is the content clearly labeled as AI-generated where required by applicable platform policy, state law, or to avoid consumer confusion?

A Privacy-Conscious Way to Test Your Own Voice

For a consented test with your own voice, TTSBox opens in your browser with nothing to install, no upload, and no signup.

TTSBox supports six languages and does not offer a batch API or real-time streaming, so teams that need broader language coverage or production-grade APIs may need a commercial service such as ElevenLabs — while still remaining responsible for consent and all applicable legal requirements.

FAQ

Is AI voice cloning legal in the U.S.?

It depends on how and for what purpose it is used. Cloning your own voice for your own non-deceptive project carries low legal risk. Cloning another person’s voice without authorization can violate state publicity rights, the Lanham Act, FTC rules, or copyright law depending on the facts. Protected uses such as satire, commentary, or news reporting may be lawful without consent in some circumstances, but each situation is fact-specific. This article describes general U.S. principles and is not legal advice.

Does a person’s voice have copyright?

A person’s voice is generally not itself a copyrightable work under U.S. federal law. What may be separately protected is a particular sound recording, an underlying script or song, or creative choices a human author made in producing an output. State publicity rights and the Lanham Act can protect voice identity independently of copyright.

Can I legally clone a celebrity’s voice?

Generally not without authorization for most commercial purposes. Midler v. Ford Motor Co. (9th Cir. 1988) and Waits v. Frito-Lay (9th Cir. 1992) established that using a recognizable voice in commercial advertising without consent can constitute voice misappropriation — even when no recording is copied. State publicity laws and the Lanham Act add further protection. Parody or satire may be protected, but only when the expression is clearly non-deceptive and meets the applicable legal standard in the relevant jurisdiction.

Is cloning my own voice always legal?

Cloning your own voice addresses the consent and publicity-rights question for voice identity. It does not resolve: the copyright in any script or song the cloned voice performs, FTC or Lanham Act issues if the output is used commercially in a misleading way, TCPA/FCC requirements if the output is telephoned to consumers, or applicable platform or contractual terms. “Cloning your own voice” is low-risk on one dimension, not automatically risk-free on all dimensions.

Does consent have to be in writing?

It does not have to be — oral consent can be valid — but written consent is strongly recommended. Oral agreements are difficult to prove, and scope is harder to establish in a dispute. Written consent that specifies the voice owner, the project, the platform, the duration, commercial use rights, modification rights, and a revocation procedure gives you documentation if questions arise later.

What would the NO FAKES Act change?

If enacted, S.4591 would create a federal right for all U.S. individuals — not only celebrities — against unauthorized AI-generated voice and visual replicas, with exceptions for news, commentary, parody, and satire (subject to specific statutory conditions). It would establish a notice-and-counter-notice process and limitations on provider liability. As of July 14, 2026, it remains a bill pending full congressional passage and has not become law.

Are AI voice robocalls legal?

Only if the recipient has given prior express consent as required by the TCPA and FCC rules. The FCC ruled in February 2024 that AI-generated voice calls are subject to the same TCPA prior-consent requirements as other robocalls. Marketing calls require prior express written consent from the specific recipient. Consent to use a person’s voice for other purposes does not satisfy this requirement. State robocall laws may add further requirements.

Can I clone the voice of someone who has died?

It depends on the state. Several states — including California, Indiana, and Tennessee — extend publicity rights post-mortem for a defined period. The estate may control both the voice identity rights and the copyright in any source recordings. Getting authorization from the estate, and verifying the applicable state’s post-mortem duration and scope, is required before proceeding.

Does fair use allow voice cloning?

Fair use is a federal copyright doctrine (17 U.S.C. § 107) that considers purpose and character of the use, nature of the original work, amount copied, and market effect. It applies to copyright claims, not to state publicity rights, the Lanham Act, FTC rules, or TCPA. A use might qualify as fair use for copyright purposes but still violate publicity law or be deceptive under FTC standards. Fair use is always fact-specific and is not a blanket permission.

Do I have to disclose that audio is AI-generated?

Not universally, but disclosure is recommended and may be required in specific contexts: (a) some state laws require disclosure of AI-generated content in defined categories such as political advertising; (b) platform policies may separately require AI content labels; (c) the July 2026 joint RIAA/IFPI labeling initiative is a voluntary framework for sound recordings, not a legal mandate; (d) disclosing AI generation can help avoid false-endorsement or consumer-deception claims even where no law expressly requires it.

Sources

• U.S. Copyright Office — Copyright and Artificial Intelligence Reports (Parts 1–3): copyright.gov/ai
• U.S. Copyright Office — Part 2 AI Report announcement (January 2025): newsnet/2025/1060
• NO FAKES Act S.4591 (119th Congress) — reported text: govinfo.gov BILLS-119s4591rs
• Senate Judiciary Committee — June 18, 2026 executive business meeting results: judiciary.senate.gov
• White House — National Policy Framework for AI Legislative Recommendations (March 2026): whitehouse.gov
• Tennessee ELVIS Act (Public Chapter 588, 2024) — Tenn. Code Ann. § 47-25-1101 et seq.
• FCC — Declaratory Ruling on AI-Generated Voice Robocalls (February 2024): FCC-24-17A1
• FTC — Government and Business Impersonation Rule (2024): ftc.gov/business-guidance/blog/2024/02
• RIAA/IFPI — AI Labeling Initiative announcement (July 10, 2026): riaa.com
• Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988)
• Waits v. Frito-Lay, Inc., 978 F.2d 1093 (9th Cir. 1992)