Voice Cloning Lawsuit 2026: Legal Cases Roundup
Voice cloning lawsuit activity in 2026 is no longer an edge case — it is mainstream intellectual property and civil rights law. From a Hollywood actor’s dispute with a major AI lab over an eerily familiar-sounding voice assistant, to a U.S. Senate bill that would give every person the right to sue over unauthorized AI voice replicas, to the EU’s first binding transparency rules for synthetic audio, the legal landscape around AI-generated voices changed faster in the past 24 months than in the previous decade. This roundup covers every major case, every significant law, and what the combined picture means for content creators, developers, and anyone who works with voice AI.
TL;DR
- The Scarlett Johansson / OpenAI “Sky” voice dispute (2024–ongoing) became the highest-profile AI voice rights case globally and directly influenced U.S. legislative momentum.
- Drake’s “Heart on My Sleeve” (2023) established that AI-cloned artist voices implicate both copyright and right-of-publicity law — record labels are pushing hard for statutes.
- The U.S. NO FAKES Act is in its second Senate iteration; it would create a federal cause of action for unauthorized AI voice replicas.
- Tennessee’s ELVIS Act is already in force — the first U.S. state law explicitly covering AI voice clones of performers.
- The EU AI Act’s transparency obligations for synthetic voice content became active August 2, 2026.
- Brazil’s LGPD treats voiceprints as sensitive biometric data; unauthorized AI voice cloning triggers enforcement by the ANPD.
- Practical implications for creators: consent, labeling, and jurisdiction-checking are no longer optional.
The Scarlett Johansson / OpenAI “Sky” Dispute
No single event did more to push AI voice rights into mainstream legal and political conversation than the dispute between Scarlett Johansson and OpenAI in May 2024 — and the fallout is still playing out in 2026.
The facts, as publicly reported: OpenAI launched GPT-4o with a set of voice assistant personas, one of which — named Sky — had a vocal quality that struck many listeners as strikingly similar to Johansson’s voice. Johansson released a public statement saying she had been approached by OpenAI CEO Sam Altman the previous September and asked to license her voice. She declined. Months later, Sky launched. Johansson said she was “shocked, angered and in disbelief” and engaged legal counsel to demand that OpenAI explain how the voice was created.
OpenAI voluntarily paused the Sky voice and said the similarity was “unintentional” — the voice was recorded with a different voice actor who naturally had a similar timbre. The company did not publicly release any documentation to verify this claim, and no settlement terms were disclosed as of mid-2026.
Why the case matters legally, even without a filed lawsuit:
- It demonstrated the gap in current law. Johansson did not have a clear federal cause of action for AI voice similarity short of recording contracts, trademark law, or California’s right-of-publicity statute — none of which map cleanly onto “an AI voice that sounds like you but is technically a different actor.”
- It accelerated legislative drafting. Several senators cited the case explicitly when re-introducing the NO FAKES Act later in 2024. The case gave legislators a name-recognition anchor for abstract legislation.
- It established public tolerance thresholds. Industry observers noted that the public reaction was not “how amazing — an AI that sounds like a celebrity.” It was near-universal discomfort. This shaped how AI labs adjusted their voice product launch playbooks.
For a broader look at how AI voice cloning technology works and where it is used legitimately, see our post on AI voice cloning for voiceover work.
Drake’s “Heart on My Sleeve” and the Music Industry Response
About a year before the Johansson case, a different kind of AI voice incident arrived from the music world. In April 2023, a track called “Heart on My Sleeve” appeared on streaming platforms — credited to a pseudonymous creator called Ghostwriter — using AI-cloned vocals that sounded convincingly like Drake and The Weeknd. The track accumulated millions of streams before Universal Music Group forced its removal on copyright grounds.
The case exposed a structural ambiguity in U.S. copyright law:
Sound recordings are protected (typically by the record label that funded the master recording). The voice itself is not separately copyrightable at the federal level. Right-of-publicity law — which does protect voice — is primarily state law, meaning enforcement and standards vary by jurisdiction.
The Recording Industry Association of America (RIAA) and the major labels used the 2023 incident to lobby aggressively for federal legislation. Their core argument: if you can train an AI on a specific artist’s voice and release commercial product with that voice, existing law provides too narrow a remedy. The 2024–2025 push for the NO FAKES Act is in large part a direct response to “Heart on My Sleeve” and the dozens of similar tracks that followed it.
For context on how the creator economy intersects with voice AI tools, see voice changer for content creators.
The NO FAKES Act: Status and Scope in 2026
What the Bill Would Do
The NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act) is federal legislation introduced in the U.S. Senate. As of mid-2026, it has gone through two iterations and is in committee markup in the Senate Judiciary Committee. The bill’s core provisions:
| Provision | Detail |
|---|---|
| Protected right | Every individual has a property right in their voice and likeness — AI replica or otherwise |
| Covered entities | Platforms that knowingly host unauthorized AI replicas; producers who create them |
| Consent requirement | Commercial use of an AI voice replica requires explicit prior written consent from the individual |
| Damages (non-commercial) | $5,000 per violation |
| Damages (commercial) | Actual damages plus disgorgement of profits |
| Death — posthumous rights | Survives 70 years post-mortem for performers (10 years for non-public figures) |
| Safe harbor | Platforms that act expeditiously on takedown notices get DMCA-style protection |
| Satire/parody exception | News commentary, satire, parody, criticism, and scholarship are explicitly excluded |
What Changed Between Drafts
The 2024 draft drew criticism from digital rights organizations — including the Electronic Frontier Foundation — for language that critics said could chill transformative creative works, fan fiction, and parody. The 2025 revised draft responded by tightening the definition of “AI replica” (requires a realistic representation that a reasonable person would mistake for the authentic individual) and expanding the satire/parody carve-out. The revised draft also removed a provision that would have imposed liability on AI model trainers, narrowing focus to deployment and distribution.
Likelihood of Passage
No voice replica bill has passed the full Congress as of this writing. The NO FAKES Act has bipartisan support in the Senate but faces resistance in the House from tech-caucus members who cite concerns about AI development chilling effects. The most likely path to passage is as part of a broader AI or intellectual property omnibus — standalone passage is considered unlikely in 2026.
Tennessee ELVIS Act: The First State Law to Target AI Voice Clones
Background
Tennessee passed the ELVIS Act (Ensuring Likeness Voice and Image Security Act) in March 2024, with an effective date of July 1, 2024. It is the first U.S. state statute to explicitly address AI-generated voice replicas of performers.
What It Covers
- Any commercial use of an AI-generated replica of a recording artist’s voice without consent is prohibited
- Applies to both recordings and live performance simulacra
- Civil damages: up to $10,000 per violation (or actual damages if higher) plus attorney’s fees
- Criminal penalties for willful commercial violations: Class A misdemeanor (up to 11 months and 29 days in jail in Tennessee)
- The protections survive death for a 10-year post-mortem window (extendable)
Why Tennessee?
Nashville is the center of the U.S. country music industry and one of the largest music markets globally. The state’s lawmakers had specific industry pressure to act after AI-cloned country artist voices appeared in unauthorized commercial releases in late 2023. The RIAA and the Country Music Association supported the bill actively.
The ELVIS Act is named partly as an homage to Elvis Presley — whose estate has historically been assertive about unauthorized commercial use of his voice and likeness — but the law covers all performers, not just the King.
Other State-Level Activity
Several other states have introduced or passed voice-replica legislation:
| State | Status | Scope |
|---|---|---|
| Tennessee | In force (July 2024) | Performers; commercial AI voice replicas |
| California | AB 2602 signed October 2024 | Contracts where AI replaces performer work |
| New York | S 5334 in committee (2026) | Right of publicity updated for AI |
| Texas | HB 4337 passed 2025 | Biometric data + voice replica |
| Illinois | BIPA (existing) | Voiceprints as biometric data; broad existing coverage |
Illinois is notable because its existing Biometric Information Privacy Act (BIPA) already covers voiceprints — courts have found that voice data used to train AI models can constitute biometric information under BIPA, meaning Illinois may have de facto AI voice clone protection even without a dedicated new statute.
EU AI Act: Voice Biometric Rules and Transparency Obligations
Overview of Relevant Provisions
The EU AI Act (Regulation (EU) 2024/1689, in force August 2024) creates a tiered risk classification for AI systems. Voice AI intersects with the regulation in two places:
1. Prohibited AI practices (Article 5) — applied from February 2025:
- AI systems for real-time remote biometric identification in public spaces are prohibited with narrow law enforcement exceptions
- “Subliminal” AI systems designed to manipulate people without their awareness are prohibited
- Voice deepfakes used to exploit psychological vulnerabilities are within scope of the manipulation prohibition
2. Transparency obligations (Article 50) — applied from August 2, 2026:
- Deployers of AI systems that generate synthetic voice audio intended for public consumption must clearly disclose that the content is AI-generated
- The disclosure must be in a machine-readable format and in a “prominent” form accessible to the recipient
- Exemptions exist for clearly artistic or clearly satirical contexts where the AI nature is already obvious
3. High-risk classification:
- AI systems used for biometric identification (including voice) of individuals are classified as high-risk and subject to conformity assessment, registration in the EU database, and human oversight requirements
What This Means for Voice AI Developers
Any product that uses AI to generate audio content — including cloned voices — and distributes it in the EU must implement:
- A technical disclosure mechanism (watermarking, metadata, or explicit on-screen/audio notice)
- Documentation demonstrating compliance with transparency requirements
- A risk management system covering potential misuse for impersonation
The enforcement body in each EU member state (e.g., the BNetzA in Germany, the Ofcom equivalent in France) can impose fines of up to €15 million or 3% of global turnover for violations of transparency obligations.
Practical Contrast: EU vs. U.S.
| Dimension | EU AI Act | U.S. (current) |
|---|---|---|
| Federal law in force | Yes (August 2026 transparency) | No (patchwork of state laws) |
| Disclosure required | Yes, prominently | Not federally; varies by state |
| Biometric voice data protection | High-risk classification + Article 50 | State law (BIPA in IL, etc.) |
| Penalties | Up to €15M or 3% global revenue | Varies; NO FAKES Act not passed |
| Satire/parody exemption | Yes, explicit | Yes, in NO FAKES Act proposal |
Brazil’s LGPD and Voice Biometric Data
Brazil’s LGPD (Lei Geral de Proteção de Dados, Law 13,709/2018) is the country’s comprehensive data protection statute, modeled partly on the EU’s GDPR. Its application to AI voice cloning is through the classification of voice data as sensitive personal data.
How LGPD Applies
Article 5, Section II defines sensitive personal data to include “biometric data,” and Brazilian courts and the ANPD (Autoridade Nacional de Proteção de Dados) have confirmed that voiceprints — the acoustic signature extracted from speech — qualify as biometric data under this definition.
Processing sensitive personal data under LGPD requires one of a narrow set of legal bases:
| Legal basis | Conditions |
|---|---|
| Explicit consent | Must be separate, informed, specific, and freely given — cannot be bundled into general terms of service |
| Legal obligation | Processing required by law |
| Vital interests | Emergency protection of life or health |
| Legitimate interest | Applies only to non-sensitive data — NOT available for biometric voiceprints |
This means that processing someone’s voice to create an AI clone — even for seemingly benign commercial uses like a branded voice assistant — requires explicit, separate, specific consent under LGPD. There is no legitimate-interest bypass for voice biometrics.
Fines: The ANPD can impose fines of up to 2% of a company’s revenue in Brazil in its most recent fiscal year, capped at R$50 million per infraction. Repeated violations and cases involving sensitive data attract the maximum end of the range.
Enforcement trend: The ANPD increased enforcement activity in 2025 and 2026, including several investigations involving tech companies processing biometric data without adequate legal basis. Voice AI is specifically mentioned in the ANPD’s 2025–2027 strategic plan as a priority enforcement area.
For creators and developers targeting Brazilian audiences: LGPD compliance means obtaining explicit consent before cloning or processing any person’s voice, maintaining documentation of that consent, and honoring data subject requests to delete voice data.
Additional Cases and Legal Developments to Watch
FTC Voice Impersonation Rule (2024)
The U.S. Federal Trade Commission finalized an update to its impersonation rule in March 2024 to explicitly cover AI-generated voice impersonation of government officials and businesses. The rule allows the FTC to seek civil penalties directly (not just injunctive relief) against companies that:
- Use AI to impersonate government agencies or officials
- Falsely claim government affiliation or endorsement through voice deepfakes
This is narrower than the NO FAKES Act (it covers institutional impersonation, not individual voice rights generally) but it is in force now and has already been cited in FTC enforcement letters to voice AI companies.
Music Licensing Negotiations: The AI Training Data Question
One underreported dimension of the legal landscape is the ongoing negotiation between AI labs and music rights holders over training data. Several major labels filed suit in 2024–2025 alleging that training voice synthesis models on recordings constitutes copyright infringement of the underlying sound recordings. As of mid-2026, these cases are in discovery; no court has ruled definitively.
The outcome matters enormously for voice AI products. If training on unlicensed audio is found to infringe copyright, the development cost of compliant voice AI systems will increase substantially — potentially concentrating the market among well-capitalized players with licensing agreements.
Right-of-Publicity Law: Increasing Scope
Even without new federal legislation, right-of-publicity law in states like California, New York, and Indiana covers the commercial use of a person’s voice. What changed in 2024–2026 is how courts interpret “commercial use” — several rulings have found that releasing AI-generated audio commercially can constitute misappropriation of voice identity even when no literal recording of the person was used, if the output is designed to sound like that person.
The “designed to sound like” standard is significant. It shifts the question from “did you sample this person’s voice?” (a factual question) to “did you intend to evoke this person’s identity?” (a purpose and intent question). This is harder to escape with technical arguments about training data provenance.
What This Means for Voice AI Creators and Developers
Whether you build voice tools, create content with AI voice technology, or just use voice software as a streamer or podcaster, the legal picture in 2026 has concrete practical implications:
For Content Creators
| Action | Why It Matters |
|---|---|
| Clone only voices you have rights to | Right-of-publicity and LGPD (in BR) create liability for cloning real people without consent |
| Add AI disclosure to published audio | EU AI Act requires it for EU audiences; increasingly expected globally |
| Do not impersonate real people commercially | FTC rule + state right-of-publicity + nascent federal law all point here |
| Check platform terms of service | Most major platforms (YouTube, TikTok, Spotify) have AI voice-impersonation policies that can trigger demonetization or removal |
| Keep consent documentation | If you have permission, document it in writing |
For Developers
If you are building a product that clones voices:
- Build a consent layer first, not as an afterthought. LGPD requires explicit, specific consent for voiceprint processing; EU AI Act requires transparency disclosures. Design for these from day one.
- Implement technical safeguards against impersonation. Several U.S. and EU legal frameworks hold platforms partially liable for “knowingly” hosting unauthorized replicas. Proactive detection and filtering reduces exposure.
- Document your training data provenance. If music-training-data lawsuits succeed, unlicensed training data becomes a liability. Maintain records now.
- Watch the NO FAKES Act markup. If it passes even in amended form, federal liability for unauthorized AI voice replicas would change the cost-benefit analysis of many product features.
Tools like VoxBooster that focus on personal voice processing — transforming your own voice in real time for streaming, gaming, or content creation — operate in a cleaner legal space than tools designed to replicate third-party voices. Understanding that distinction helps creators make responsible choices. For more on how voice AI technology is used legitimately in creative work, see AI voice cloning for voiceover production and our data overview in voice cloning statistics 2026.
Scam and Fraud Liability: A Separate Legal Track
The legal cases above concern creative and commercial rights. There is a parallel and faster-moving legal track focused on fraud and scam use of cloned voices — and enforcement here is already active.
Voice cloning has been used in documented cases of:
- Grandparent scams — AI voice impersonating a grandchild calling in an emergency
- CEO fraud — CFOs wiring money after hearing a cloned executive voice
- Election interference — robocalls using cloned candidate voices (the FCC made AI voice robocalls illegal in February 2024)
Federal wire fraud and identity theft statutes already apply to these uses without any new legislation. Several state AGs have brought cases under existing fraud law against individuals using voice AI for scams. For a detailed training framework on recognizing and resisting these attacks, see voice cloning scam awareness training.
FAQ
Is AI voice cloning legal in the United States?
It depends on how the voice is used. Cloning your own voice or creating fictional characters is generally legal. Cloning a real person’s voice without consent — especially for commercial use, impersonation, or to create false impressions — can violate right-of-publicity statutes, the Lanham Act, or, if the NO FAKES Act passes, federal law. Several states already have specific protections.
What is the NO FAKES Act and what does it do?
The NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act) is a proposed U.S. federal law that would create an individual right to control AI-generated replicas of one’s voice and likeness. It would allow affected individuals to sue platforms that knowingly host unauthorized AI replicas and establish damages including $5,000 per violation for non-commercial infringement.
What did the ELVIS Act change for voice cloning in Tennessee?
Tennessee’s ELVIS Act (Ensuring Likeness Voice and Image Security), effective July 1, 2024, updated the state’s right-of-publicity law to explicitly cover AI-generated voice replicas. It makes it a civil and criminal violation to use an AI-generated replica of a performer’s voice for commercial purposes without consent, with statutory damages up to $10,000 per violation.
Does the EU AI Act regulate voice cloning?
Yes. The EU AI Act classifies systems that generate synthetic voice biometrics — especially for impersonation — as high-risk AI in certain contexts and requires transparency labeling for AI-generated audio content intended for public consumption. The transparency obligations under Article 50 applied from August 2, 2026. Biometric identification systems face stricter prohibited-use rules from February 2025.
How does Brazil’s LGPD apply to AI voice cloning?
Brazil’s LGPD classifies voiceprint data as sensitive personal data under Article 5. Processing someone’s voice for AI cloning purposes requires explicit, informed consent. Unauthorized commercial use of a voice replica can trigger LGPD enforcement by the ANPD, with fines up to 2% of Brazilian revenue, capped at R$50 million per infraction.
What happened between Scarlett Johansson and OpenAI over the ‘Sky’ voice?
In May 2024, OpenAI launched a voice assistant called Sky that many listeners — including Johansson herself — described as sounding nearly identical to her voice. Johansson stated she had declined a licensing offer from OpenAI CEO Sam Altman and was shocked when the company launched a voice with her distinctive timbre anyway. OpenAI paused Sky and said the resemblance was unintentional. No settlement details were publicly confirmed as of mid-2026, but the dispute renewed calls for federal voice-replica legislation.
What should creators know before using AI voice cloning software?
Creators should verify they have rights to the source audio (their own voice or licensed voice with appropriate permissions), not use the output to impersonate real people without consent, comply with any platform terms of service, add transparency disclosures if publishing AI-generated audio publicly, and check local right-of-publicity or biometric privacy laws in their state or country. When in doubt, consult an entertainment or IP attorney.
Conclusion
Voice cloning lawsuit activity in 2026 reflects a legal system catching up to a technology that moved faster than anyone anticipated. The Scarlett Johansson / OpenAI dispute provided a public focal point. The Drake / “Heart on My Sleeve” incident gave the music industry a lobbying anchor. The NO FAKES Act is the most significant pending federal response. Tennessee’s ELVIS Act showed that states can and will act ahead of Congress. The EU AI Act is already in force. Brazil’s LGPD already covers voiceprints as sensitive biometric data.
The practical bottom line for anyone working with voice AI: consent documentation, transparency disclosures, and jurisdiction awareness are the three pillars of responsible use. The legal frameworks are still evolving — the next 12 months will likely bring at least one major court ruling and possibly federal legislation — but the direction of travel is clear. Unauthorized use of real people’s voices is becoming measurably riskier from a legal standpoint in every major market simultaneously.
Tools that focus on processing your own voice — real-time effects, personal voice customization, creative persona building — remain in clearly legitimate territory. VoxBooster is built around that use case: modifying your own voice for streaming, gaming, and content creation, with no third-party voice database and no mechanism for impersonating real people. If you are exploring what responsible voice AI looks like in practice, the ai voice generator market outlook for 2027 lays out where the industry is heading under the emerging regulatory pressure. You can also download VoxBooster and try the 3-day free trial to see how voice AI tools work when built for personal creative use rather than impersonation.