Voice Cloning and the Law: Copyright & Ethics in 2026

In a hypothetical fall 2025 scenario, an AI cover song using an unauthorized clone of a famous rapper's voice surpassed one million views on YouTube. The original artist threatened legal action, but the uploader claimed "AI made it" and argued it wasn't copyright infringement. Around the same time, a Korean dubbing studio used AI to recreate the voice of a deceased voice actor to complete an unfinished animation, sparking heated debate among the family and fans. Voice cloning technology is now routine for creators, dubbing studios, and education platforms—but the legal and ethical boundaries remain murky. As of 2026, the global music and content industries are deep in copyright disputes and self-regulatory guideline development around AI voice clones. This article breaks down where voice cloning is legal, when it crosses ethical lines, and the practical guidelines content creators need to know right now.

Key Takeaways

Voice cloning itself is legal, but commercial use without explicit consent from the original speaker creates significant litigation risk for violation of right of publicity.

Under Korean copyright law, voice is not a copyrighted work, but performers' moral and economic rights are protected. Cloning a deceased person's voice requires family consent to avoid legal disputes.

U.S. government reports and the Congressional Research Service (CRS) recognize that AI-generated voice and video replicas can infringe copyright and right of publicity, and recommend federal legislation against unauthorized digital replication.

As of 2026, major AI voice platforms worldwide have adopted "authorized voice synthesis" principles as self-regulation, and contracts specifying scope, duration, and revenue sharing have become industry standard.

To safely use voice clones in dubbing, audiobooks, or educational content, obtain written consent from the original speaker and specify usage purpose, languages, and distribution channels in the contract.

Is Voice Legally a "Copyrighted Work"?

Voice itself is not protected as a copyrighted work under Korean copyright law. Copyright law protects "creative works expressing ideas or emotions," and voice is merely a means of expression, not a creation in itself. But performances containing voice are different. When a voice actor performs dialogue or a singer performs a song, that performance is protected and generates performers' rights under copyright law. Therefore, using AI to clone someone's voice and generate new dialogue can be considered unauthorized use of the original performer's performance.

More importantly, right of publicity comes into play. Voice is a unique characteristic that identifies an individual, so commercial use without consent constitutes infringement of right of publicity. Especially for famous actors, voice actors, and singers, whose voices have significant economic value, right of publicity (the economic right to use personal identifiers like name and likeness) are recognized. In 2024 in the United States, an AI cover song using an unauthorized clone of rapper The Weeknd's voice sparked controversy, and the Recording Industry Association of America (RIAA) filed lawsuits claiming AI platforms' use of voice data for training without artist consent was copyright infringement.

On this issue, the U.S. Copyright Office issued a report acknowledging the potential for copyright and right of publicity infringement by AI-generated voices, images, and other digital replicas, and recommended federal legislation against distribution of unauthorized replicas. The report broadly defines "voice" to include both actual and simulated voices, and is becoming the standard for future AI voice clone regulation. [U.S. Copyright Office, Copyright and Artificial Intelligence Part 1: Digital Replicas Report](https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf)

Similar cases are increasing in Korea. At a 2025 National Assembly press conference, critics pointed out that in the absence of social consensus or legislative conclusions on AI music copyright issues, it was inappropriate for certain organizations to set standards through their own regulations. While industry self-regulation is proceeding in the legal vacuum, the lack of clear legal basis means low predictability in disputes.

Can You Clone a Deceased Person's Voice with AI?

Cloning deceased persons' voices is the most legally and ethically sensitive area. An ethics expert at IAAE stated: "For AI voice cloning, consent from the voice owner is appropriate. If consent cannot be obtained (deceased persons), family consent is necessary."

Under Korean civil law, personality rights expire upon death, but criminal defamation of the deceased is criminally punishable. Additionally, families have certain protective rights over the deceased's personality markers like likeness and voice. Therefore, to commercially use AI to clone a deceased person's voice, prior consultation with and written consent from the family is essential.

In practice, the following must be specified in contracts:

Purpose of cloned voice use (e.g., completing unfinished animation, audiobook narration)

Duration and territory (e.g., worldwide perpetual use vs. limited period)

Revenue sharing terms (whether to pay family a percentage)

Prohibition on voice alteration/distortion (to avoid defaming the deceased)

This practice is already established overseas. In the 2023 film "Top Gun: Maverick," deceased actor Val Kilmer's voice was recreated with AI, but the production company obtained prior consent from him and his family before proceeding with voice cloning. Conversely, cases proceeding without consent have mostly led to legal disputes.

What Data Do AI Voice Platforms Train On?

The core of voice cloning technology is training data. Whose voice the AI model learned from and through what consent process determines legal risk. According to U.S. Copyright Office and Congressional Research Service reports, AI-generated outputs like voice and video cannot be considered human works if the AI determined the specific elements of expression; only parts where humans intervened with creative choices, arrangement, and editing receive limited protection. In other words, human polishing (editing, modification) in AI voice dubbing and clone content can generate copyright. [U.S. Copyright Office, AI Copyright Report Part 2](https://www.copyright.gov/policy/artificial-intelligence/)

Additionally, the Congressional Research Service (CRS) clarifies the legal difference between using AI as an assistive tool versus replacing creation, noting that only human contributions are subject to copyright protection, and AI-generated portions must be identified and excluded during registration. This serves as direct reference material for copyright and contract design in B2B AI dubbing and audiobook services. [Congressional Research Service, Generative AI and Copyright Law Analysis](https://www.congress.gov/crs-product/LSB10922)

As of 2026, major AI voice platforms have adopted "authorized voice synthesis" principles as self-regulation. This means using only data for which voice providers have explicitly consented for training, and specifying consent scope (purpose, duration, territory) in contracts. However, some free open-source tools collect voices from public YouTube videos or podcasts without permission for training. Using such tools in commercial projects makes defense difficult if the original speaker later files suit.

How to Safely Use Voice Clones in Dubbing and Audiobook Production

For content production companies and creators to safely use voice clones, follow this checklist:

Stage	Required Action	Legal Basis
1. Original speaker engagement	Written consent covering purpose, duration, territory, and revenue terms	Right of publicity
2. Training data acquisition	Use only voice recorded by speaker (no unauthorized collection from YouTube, etc.)	Performers' rights under copyright law
3. Contract drafting	Specific scope of voice use (languages, channels, duration), include NDA	Civil law (contract)
4. Platform selection	Prioritize B2B platforms with stated authorized voice synthesis principles	Risk management
5. Credit attribution	Credit original speaker (e.g., "Voice provided by: ○○○, AI clone technology applied")	Respect for right of publicity

Contract drafting is especially essential. Verbal agreements alone make proof difficult if disputes arise later. Contracts must include at minimum:

Purpose of use: "Limited to Korean, English, and Japanese dubbing of animation ○○○"

Duration: "From January 1, 2026 to December 31, 2028" or "perpetual use"

Territory: "Worldwide" or "limited to Korea, United States, Japan"

Revenue sharing: "○% of net revenue" or "a one-time fee in local currency"

Prohibited uses: "Voice may not be used for political, religious, or adult content"

Termination conditions: "Use must cease within 30 days of written request by speaker"

NDAs (non-disclosure agreements) should also be executed. When recording unreleased animation, drama, or game dialogue, legally bind voice actors or AI platforms not to leak scripts externally.

How Far Has Industry Self-Regulation Come?

As of 2026, industry self-regulation is rapidly expanding to fill the legal vacuum around voice cloning technology. The U.S. Copyright Office and Recording Industry Association of America (RIAA) are demanding transparency in AI training data sources and pursuing class action lawsuits against unauthorized training platforms. The European Union mandated disclosure of training data used by generative AI through the AI Act, and voice cloning platforms are included in this regulation. [U.S. Copyright Office, Copyright and Artificial Intelligence Part 1: Digital Replicas Report](https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf)

Korea does not yet have clear legislation, but the Korea Communications Commission (KCC) and Ministry of Culture, Sports and Tourism are preparing generative AI guidelines. At a 2025 National Assembly press conference, critics noted the need for social consensus on AI music copyright issues and criticized certain organizations for inappropriately setting standards through their own regulations.

Within the industry, the "Ethical AI Voice Pledge" is spreading. Major AI voice platforms are voluntarily adopting these principles:

Do not use voice providers' data for training without explicit consent.

Specify consent scope (purpose, duration, territory) in contracts.

Allow voice providers to withdraw consent and request data deletion at any time.

Monitor to ensure generated voices do not defame original speakers or get used in illegal content.

These regulations and guidelines are contributing to increased trust and transparency in the AI voice cloning market and reducing legal risk for both creators and platforms.

---

Voice cloning technology has revolutionized content production speed and cost efficiency, but failing to observe legal and ethical boundaries creates significant risk. As of 2026, in the absence of clear legislation, production companies and creators must protect themselves with contracts, consent forms, and NDAs, and choosing platforms that follow authorized voice synthesis principles is the best risk management. AI voice regulation will become more specific going forward, and ethical practices established now are likely to become industry standards.

Polita is a B2B dubbing and audiobook service combining AI dubbing with human polishing. While leveraging voice clone technology, Polita operates with original actor consent and NDA-based security workflows as standard. If you need dubbing that minimizes legal and ethical risk while preserving the original's soul, you can safely execute projects with partners like Polita.

Frequently Asked Questions

Q. Is it illegal to clone a famous voice actor's voice with AI for dubbing?

Commercial use without explicit written consent from the individual creates significant litigation risk for right of publicity infringement, and can also infringe performers' rights under copyright law. You must draft a contract specifying purpose, duration, and revenue sharing.

Q. Can I use publicly available YouTube voices for AI training?

Even for public content, unauthorized use as training data can constitute copyright and right of publicity infringement. As of 2026, the U.S. Copyright Office and RIAA are pursuing class action lawsuits against unauthorized training platforms, and the EU AI Act mandates disclosure of training data sources.

Written consent from the family is essential. While the deceased's personality rights expire upon death, criminal defamation of the deceased is criminally punishable, and families have certain protective rights over the deceased's likeness and voice. Purpose, duration, and revenue sharing must be specified in the contract.

Q. What should I check when choosing an AI voice platform?

Verify whether they state "authorized voice synthesis" principles, transparently disclose training data sources, and support NDA and contract drafting. B2B platforms tend to have better-established legal risk management systems.

Q. What items must be included in a voice clone usage contract?

You must specify: purpose of use (content name, application), duration (start/end dates or perpetual), territory (worldwide or specific countries), revenue sharing terms, prohibited uses (no political or adult content), and termination conditions (cessation procedure upon speaker request).

---

References

- U.S. Copyright Office, Copyright and Artificial Intelligence Part 1: Digital Replicas Report — The U.S. Copyright Office's first official analysis of copyright, right of publicity, and the need for digital replica regulation for AI-generated voice and video, recommending federal legislation against unauthorized replicas. - U.S. Copyright Office, AI Copyright Report Part 2 — Clarifies copyrightability of AI-generated outputs and the standard that only parts with human creative intervention are protected. Basis for copyright arising from human contributions like polishing in AI dubbing and cloning. - Congressional Research Service, Generative AI and Copyright Law Analysis — The Congressional Research Service explains the legal difference between AI as assistive tool versus creative replacement, and that only human contributions are subject to copyright protection. Reference material for B2B AI dubbing and audiobook contract design.