Celebrity Voice Clone Legal: What AI Impersonation Law Allows
Celebrity voice clone legal questions are exploding right now — and for understandable reasons. The same AI tools that let a streamer sound like their favorite character can produce audio indistinguishable from a real person’s voice. Legislators noticed. Between 2023 and 2025, the US and EU passed more voice-related AI law than in the previous two decades combined. This guide maps exactly what the law says, where the lines sit, and what you can safely do with AI voice impersonation technology as a creative, a fan, or a developer.
Key Takeaways
- The Tennessee ELVIS Act (2024) is the first US law targeting AI voice simulation specifically — it extends right-of-publicity to voice in Tennessee and applies to any commercial use without consent.
- California AB 2602 (2023) voids contracts that use AI voice replicas of performers without explicit written consent.
- California AB 1836 (2025) protects deceased celebrities’ voices via estate rights for 70 years.
- The EU AI Act mandates disclosure labels on synthetic audio and prohibits deceptive AI impersonation.
- Parody and satire of public figures retain robust First Amendment protection — the key test is whether a reasonable viewer could be deceived.
- SAG-AFTRA’s Replica Voice provisions require informed consent and residual payments for AI versions of union performers.
- Private creative use and clearly labeled fan content carry the lowest legal risk.
Why Celebrity Voice Clone Law Moved So Fast
Before 2020, most right-of-publicity statutes focused on image: a celebrity’s face in an advertisement, their likeness on a product. Voice had secondary protection in most states. Then real-time AI voice conversion became accessible enough that you could produce a convincing simulation of a known public figure in minutes on consumer hardware.
The inflection point was a combination of factors: deepfake political content in the 2024 election cycle, voice-scam fraud targeting elderly people using cloned relative voices, and SAG-AFTRA’s prolonged strike action in which AI voice replacement clauses were a core dispute. Legislators in multiple jurisdictions responded within 18 months of each other — an unusually fast legislative cycle for technology law.
Understanding the current landscape means reading five overlapping legal frameworks at once: right of publicity (state law), the ELVIS Act (Tennessee), California’s two performer protection bills, the EU AI Act, and First Amendment / fair use doctrine. None of them contradict each other cleanly.
Right of Publicity: The Foundation
Right of publicity is a state-level legal doctrine in the US that gives individuals — including celebrities — the right to control commercial use of their name, image, likeness, and, increasingly, voice. It is distinct from copyright (which protects creative works) and trademark (which protects brand identity). Right of publicity protects identity itself.
Which states have it: California, New York, Florida, Tennessee, Indiana, and about 30 others. Scope varies significantly. California and New York have the broadest protections and are the jurisdictions most relevant to entertainment industry disputes.
What it covers: The traditional formulation covers name, signature, photograph, and likeness in commercial contexts. Post-AI-era amendments in California and Tennessee explicitly add “voice” and “AI-generated replica” language.
Duration: Most states protect living individuals. California extends protection for 70 years after death (AB 1836 strengthened this for AI specifically). New York passed similar posthumous legislation in 2021.
Enforcement: Celebrities sue under right-of-publicity statutes when AI-generated content uses their voice or likeness to imply endorsement, promote a product, or generate revenue without consent. Damages can include actual damages, profits attributable to the violation, and in some jurisdictions punitive damages.
The Tennessee ELVIS Act (2024): Voice-First Legislation
The ELVIS Act — Ensuring Likeness Voice and Image Security — was signed into law in Tennessee in March 2024 and took effect July 1, 2024. It is the first US statute designed specifically around AI voice simulation.
What it prohibits: Using an AI or other technology to produce a realistic simulation of an identifiable individual’s voice for commercial purposes without their consent. The key word is “commercial” — private non-commercial use is not covered.
Who it protects: Any individual, not just celebrities. The bill’s name references Elvis Presley (Tennessee’s most famous musical export), but the statute applies to any person whose voice is simulated.
Enforcement mechanism: It is a civil cause of action — the affected person or their estate can sue in Tennessee courts. The bill does not create criminal liability but allows for injunctive relief plus damages.
Why it matters beyond Tennessee: Tennessee is home to a significant fraction of the US music industry (Nashville). Contracts routinely specify Tennessee law. The bill has influenced similar legislation in other states, and it explicitly references AI as the primary concern — an unusual degree of technical specificity for state law.
What it does not cover: Political commentary, parody, satire, and news reporting are carved out explicitly. A comedian doing an AI-assisted impression of a musician for a sketch is not violating the ELVIS Act as long as it is not a commercial product designed to pass as real.
California AB 2602: Performer Consent in AI Contracts
California AB 2602, signed in September 2023, targets a specific scenario: what happens when a studio or production company wants to use an AI-generated version of a performer’s voice in future work.
The core rule: Any contract that purports to use an AI-generated replica of a performer’s voice must include explicit, informed consent from that performer. The consent must specify the intended use. A general contract clause that allows “any use of likeness” does not satisfy the requirement — the AI voice replica use must be spelled out.
Contracts signed before 2024: AB 2602 has a prospective effect — it does not automatically void existing contracts. However, it changes the baseline expectation for all new agreements and has been used as leverage in renegotiations.
Voice actors and narrators: The bill applies beyond on-screen talent to voice-only performers, audiobook narrators, video game voice actors, and any other performer whose voice is professionally recorded. This is significant because pre-AB 2602, many audiobook and game VO contracts contained broad rights-assignment language that studios argued covered AI replicas.
Practical impact: Studios and game developers must now have separate, explicit conversations with performers about AI voice rights. The bill effectively created a new bargaining chip in VO negotiations.
California AB 1836: Protecting Deceased Performers
AB 1836, effective January 2025, fills a gap in posthumous protection: it explicitly extends right-of-publicity claims to AI-generated likenesses and voices of deceased performers under California law.
Duration: 70 years after death, matching AB 2602’s analogous provision and aligning with federal copyright term lengths.
Who controls the right: The deceased performer’s estate, heirs, or successor-in-interest. Estates must register the claim with the California Secretary of State to enforce it.
Commercial use threshold: The prohibition applies when an AI-generated voice replica is used “in a manner designed to create the impression of a genuine performance” for commercial benefit. Archival and educational use has carve-outs.
The Elvis estate analogy: Even though Elvis Presley died in 1977, an AI voice simulation of Elvis used in a commercial without estate consent would now violate both AB 1836 (if California law applies) and the ELVIS Act (if Tennessee law applies). That is double exposure for a significant class of marketing content.
Comparison: Key US Statutes on AI Voice Cloning
| Statute | Jurisdiction | Signed | Covers Living | Covers Deceased | Commercial Threshold | Parody Exemption |
|---|---|---|---|---|---|---|
| Right of Publicity (CA Civil Code 3344) | California | 1971 (amended 2024) | Yes | Yes (70 yr post-death) | Yes | Yes |
| Right of Publicity (NY Civil Rights Law 50-51) | New York | 1903 (amended 2021) | Yes | Yes (40 yr post-death) | Yes | Limited |
| ELVIS Act | Tennessee | 2024 | Yes | Yes (via estate) | Yes | Yes |
| AB 2602 | California | 2023 | Yes (performers) | N/A | Contract context | N/A |
| AB 1836 | California | 2024 (eff. 2025) | N/A | Yes (70 yr) | Yes | Yes |
EU AI Act: Disclosure and Prohibited Practices
The EU AI Act, which entered force in August 2024 with phased implementation through 2026, approaches AI voice impersonation from a systemic risk and consumer protection angle rather than an individual rights angle.
Articles 50–52 — Transparency obligations: Any AI system that generates synthetic audio of a natural person must ensure the output is labeled as AI-generated in a manner that is “clear and distinguishable.” This applies to apps, platforms, and services — not just broadcasters. A podcast tool that uses an AI voice must label its output.
Prohibited AI practices (Article 5): AI systems “designed to deploy subliminal techniques” that deceive users without their awareness, and systems that exploit “vulnerabilities of specific groups of persons,” are prohibited. Realistic voice impersonation designed to deceive falls squarely in this category.
High-risk classification: Systems used for authentication or identity verification that involve voice are classified as high-risk under Annex III, requiring conformity assessments and human oversight mechanisms.
Enforcement timeline:
- February 2025: Prohibited practices provisions in force
- August 2025: GPAI model obligations in force
- August 2026: High-risk system obligations fully in force
What the EU AI Act does not cover: Clearly labeled entertainment, satire, or fan-made content that cannot be mistaken for genuine statements. The regulation is disclosure-focused, not creation-prohibiting. You can make AI voice impersonation content in the EU; you must label it.
Parody, Satire, and First Amendment Protection
The strongest legal protection available for AI celebrity voice content in the US is the First Amendment doctrine of parody and satire. Courts have protected voice impersonation — from radio comedians to SNL impressions — for decades. The question is whether AI-generated impersonation changes the analysis.
The core test: A two-part inquiry. First, does the work comment on or criticize the public figure being impersonated? Second, could a reasonable person mistake it for the public figure’s actual statement? If commentary is present and deception is absent, protection is robust.
Relevant precedent: Hustler Magazine v. Falwell (1988) established that parody of public figures — even offensive parody — is protected speech. The recent Ninth Circuit decisions on right-of-publicity versus First Amendment (including the “Transformative Use” test from Comedy III Productions v. Gary Saderup) provide a framework courts apply: if the creative work is primarily transformative rather than primarily a commercial exploitation of the celebrity’s identity, it is protected.
What changes with AI: The threshold question of “could a reasonable person be deceived” becomes more contested when AI synthesis produces audio with fewer tell-tale artifacts than human impressions. Explicit labeling (“This is an AI-generated parody”) shifts the analysis strongly toward protection.
Political speech: Political commentary and satire of politicians receives the highest First Amendment protection. An AI voice parody of a politician saying ridiculous things in the style of political comedy is extremely well-protected — especially on platforms that don’t pretend the audio is real.
For a deeper look at the intersection of AI voice and political content, see our post on voice cloning and political deepfake prevention.
SAG-AFTRA Replica Voice Provisions
The 2023 SAG-AFTRA strike ended with a contract that included landmark AI provisions, with replica voice clauses that created industry standards beyond just union productions.
Informed consent requirement: A signatory employer (any studio or production company that has signed SAG-AFTRA agreements) must obtain separate, informed written consent before creating an AI replica of a performer’s voice. “Informed” means the performer knows what specific project the replica will be used on.
Residuals: AI voice replicas trigger residual payment obligations comparable to actual performance residuals. You cannot create an AI replica of a union actor and then pay them once while using the replica indefinitely for free.
Scope of use: The consent must specify scope. Consent to use a replica for one animated series does not extend to other productions, promotional material, or derivative works.
Non-union implications: SAG-AFTRA contract terms are not legally binding on non-union employers. However, they have become the de facto standard for what “good practice” looks like, and they inform how courts assess whether an AI voice use was reasonable. Studios using non-union voice actors are increasingly including similar language in their contracts to avoid future liability exposure.
For a broader look at how voice cloning is used legitimately in professional contexts, see our posts on voice cloning for voiceover work and voice cloning ethics in 2026.
What Is Actually Allowed: A Practical Framework
The legal landscape is not a blanket prohibition on AI celebrity voice content. Here is a practical breakdown by use case:
Low Risk: Protected Creative Uses
- Clearly labeled parody of public figures (politicians, executives, celebrities) for commentary or entertainment purposes
- Fan-made, non-commercial content that does not claim to be a real statement
- Private use for learning, practice, or personal entertainment not distributed commercially
- Academic and journalistic analysis using brief clips as reference material
- Educational demonstrations of how AI voice cloning works, using publicly available samples
Moderate Risk: Context-Dependent
- Monetized fan content on streaming platforms — platforms’ own policies matter, and monetization increases scrutiny; consult platform ToS
- Satirical commercial products (books, merch) that reference a celebrity voice — the transformative use test applies
- Archival or memorial tribute content using a deceased person’s voice — fine with estate consent, legally uncertain without it
High Risk: Avoid Without Consent
- Commercial advertising using a celebrity voice to imply product endorsement
- Synthetic statements presented as real that could deceive an ordinary listener
- Fraud, scams, or social engineering using cloned voices — federal wire fraud laws apply independently of IP law
- Content designed to harm the reputation of a specific person (defamation law applies to AI-generated content)
- Streaming services or platforms distributing AI celebrity voice content at scale without consent
Consent Best Practices
| Scenario | Recommended Action |
|---|---|
| Using a public figure’s voice for clearly labeled satire | Label prominently; no explicit consent required |
| Generating AI voice for commercial project | Written consent from person + their agent |
| AI replica of union performer | SAG-AFTRA-compliant consent + residuals |
| AI replica of deceased celebrity | Estate written consent; verify CA/TN law applicability |
| Fan content on YouTube/Twitch | Check platform AI content policy; label clearly |
How AI Voice Detection Tools Affect the Legal Picture
One emerging factor in AI voice legal disputes is the availability of detection technology. Watermarking systems embedded at the generation layer (C2PA provenance standards, AudioSeal, and similar) can prove the origin of a voice clip in litigation.
If an AI voice tool watermarks its output and a complainant can prove a particular clip was generated by a specific tool, the legal chain of evidence becomes cleaner. Conversely, if someone strips watermarks, that act itself may constitute additional legal exposure.
The practical implication for users: tools that implement C2PA provenance watermarking shift some legal risk onto the generation event rather than the distribution event. For an overview of how these tools work, see our post on AI voice detection tools and watermarking.
Understanding detection also matters for the research side — if you are curious how AI voice models perceive similarity, voice cloning and twin research covers the biometric science behind voice individuality.
International Variations Worth Knowing
Outside the US and EU, the legal picture varies significantly:
United Kingdom: The UK does not have a federal right-of-publicity statute. Voice impersonation claims typically run through passing off (misrepresenting a commercial connection), defamation, or data protection law (GDPR via UK GDPR post-Brexit). A realistic AI voice that misrepresents a commercial endorsement can support a passing-off claim.
Canada: Right of publicity exists at the provincial level, not federally. Ontario and British Columbia have the most developed protections. Canadian courts apply the “appropriation of personality” doctrine.
Australia: Australia has limited right-of-publicity protection. Claims typically run through passing off or through the Trademarks Act if a celebrity’s voice has achieved secondary meaning as a trademark.
Japan and South Korea: Both have active legislation and court decisions addressing idol and celebrity voice synthetic content, with specific provisions relating to the entertainment industry’s common practice of extended exclusive contracts with performers.
Frequently Asked Questions
Is it legal to clone a celebrity’s voice with AI?
It depends on the purpose and jurisdiction. Cloning a celebrity’s voice for private creative use or clearly labeled parody generally has strong legal protection under the First Amendment. Commercial use, fraud, defamation, or releasing audio designed to deceive audiences can violate right-of-publicity statutes, the Tennessee ELVIS Act, or California AB 2602.
What is the Tennessee ELVIS Act?
The ELVIS Act (Ensuring Likeness Voice and Image Security), signed in 2024, makes it a civil violation in Tennessee to use AI to produce a realistic simulation of any person’s voice for commercial purposes without consent. It extends right-of-publicity protections specifically to voice, filling a gap left by older image-focused statutes.
Does the First Amendment protect AI voice impersonation parody?
Courts have consistently protected parody and satire of public figures under the First Amendment, including voice impersonation on SNL and in political commentary. The key test is whether a reasonable person could confuse the parody for the real person’s genuine statement. Clear labeling removes most legal risk.
What does California AB 2602 say about AI voice cloning?
AB 2602, signed in 2023, requires that contracts using a performer’s AI-generated voice include explicit consent clauses. Any work produced by an AI replica of a performer’s voice is void without that written consent. It strengthens the negotiating position of voice actors and performers when studios want AI versions of their voices.
What are the EU AI Act rules on deepfake voice content?
The EU AI Act classifies synthetic voice content that could deceive consumers as high-risk or prohibited depending on context. Articles 50–52 mandate clear disclosure labeling on AI-generated audio and video. Systems designed specifically to impersonate real people without consent fall under prohibited practices starting 2025.
Can I use a celebrity voice clone for a fan video or gaming stream?
Fan content and streaming commentary on public figures typically fall under fair use when non-commercial, clearly attributed, and not designed to mislead. Using a cloned celebrity voice to add jokes during a gameplay stream is almost universally treated as parody. Monetizing it commercially or presenting it as a real statement shifts the risk.
What is California AB 1836 and how does it affect deceased celebrities?
AB 1836, effective 2025, extends California’s right of publicity to AI-generated likenesses and voices of deceased performers for 70 years after death. Estates can sue anyone who uses an AI voice clone of a deceased celebrity for commercial purposes without written consent from the rights holder.
Conclusion
Celebrity voice clone legal frameworks moved from scattered state statutes to a layered federal-and-international patchwork in under three years. The pattern across all these laws is consistent: commercial deception is prohibited, clear creative expression is protected. If you are using AI voice cloning for personal experimentation, labeled satire, fan content, or professional voiceover work with proper consent, you are operating in well-protected territory. If you are deploying AI celebrity voice content commercially without consent, or in ways that could deceive an ordinary listener about the source, the ELVIS Act, California’s AB 2602 and AB 1836, and the EU AI Act all represent real legal exposure.
For tools that take responsible AI voice generation seriously — with local processing that keeps your audio on your own machine rather than on cloud servers — VoxBooster offers AI voice cloning designed for legitimate creative and professional use. The 3-day free trial requires no credit card, and everything processes locally on Windows 10/11.
Download VoxBooster — free 3-day trial, local AI processing, no credit card required.