AI voice cloning is about to get its first real test in a Japanese courtroom. On Wednesday 30 September 2026, the Tokyo District Court is due to rule on a lawsuit brought by Kenjiro Tsuda, the 55-year-old voice actor behind Kento Nanami in Jujutsu Kaisen and Seto Kaiba in Yu-Gi-Oh! Duel Monsters. He wants TikTok to take down a run of videos narrated by a deep, “lustrous” voice that he says can only be his, generated without his consent.
TikTok’s answer, according to court records seen by AFP, is that the narration is a “generic male voice” and that any resemblance to Tsuda is subjective at best. The case is believed to be the first in Japan in which a performer has asked a court to protect his vocal identity against copies made with generative AI. It comes less than two months after Japan’s justice ministry published guidelines saying a voice is protected under existing rights.
This article covers what happened, how the AI voice cloning dispute reached court, TikTok’s defence, the legal tests the judges will apply, how the ruling could go and what it means for voice actors, platforms and businesses that use synthetic voices. Our earlier look at TikTok’s AI likeness detection tool covers the platform’s own detection efforts in more depth.
Table of contents
- What Happened in the Kenjiro Tsuda AI Voice Cloning Case
- Timeline of the AI Voice Cloning Lawsuit Against TikTok
- The Money Behind the AI Voice Cloning Account
- How TikTok Is Defending Itself Against the AI Voice Cloning Claim
- Japan’s Legal Tests for AI Voice Cloning
- What the Justice Ministry’s AI Voice Cloning Guideline Says
- How the Tokyo Court Could Rule on AI Voice Cloning
- Why Voice Actors Call AI Voice Cloning an Existential Threat
- How Other Countries Treat AI Voice Cloning
- TikTok’s Own Rules on AI Voice Cloning
- What the AI Voice Cloning Case Means for Creators, Platforms and Brands
- What Comes After the AI Voice Cloning Verdict
- AI Voice Cloning and the Tsuda Case: Frequently Asked Questions
- References and Further Reading
What Happened in the Kenjiro Tsuda AI Voice Cloning Case
The facts of the AI voice cloning dispute are simple. The legal questions are not. An anonymous TikTok account posted short videos about urban legends, the occult and conspiracy theories, each narrated by a low, resonant male voice. Viewers heard Tsuda in it, and so did Tsuda.
Who Kenjiro Tsuda is
Tsuda was born on 11 June 1971 and has worked in anime since the 1990s. His best-known roles include Seto Kaiba in Yu-Gi-Oh! Duel Monsters, Overhaul in My Hero Academia and Kento Nanami, the stoic sorcerer in Jujutsu Kaisen. He also narrates, dubs foreign films and acts on screen. In Japan, fans often follow voice actors as closely as the characters they play.
That popularity is central to the AI voice cloning claim. Tatsuhiro Ueno, a professor of intellectual property at Waseda University, told AFP that the Jujutsu Kaisen role had made Tsuda’s voice “highly recognisable in Europe”. He expects the verdict to “garner a lot of international interest”.
The account at the heart of the dispute
The videos came from an anonymous account whose profile picture resembled a Jujutsu Kaisen character voiced by Tsuda. Japanese press reports of the complaint say 188 videos were posted between July 2024 and September 2025. Each paired still images with narration that appeared to be AI-generated.
Comments under the clips included remarks such as “it sounds like Tsudaken”, using his fan nickname. Tsuda’s side analysed the audio and says it found a high degree of similarity to his low, rich voice. The account can no longer be viewed on TikTok, but the videos are still at the centre of the case.
What Tsuda is asking the court to do
Tsuda is not suing the anonymous poster, because he has never been able to identify them. He is suing TikTok’s operator and asking the court to order the videos deleted. The Tokyo District Court will decide whether TikTok was responsible for removing the posts, and whether AI voice cloning of this kind infringes his rights at all.
That framing matters. A ruling against TikTok would suggest that a performer can go straight to the platform when an AI voice cloning account is anonymous. It would not need the uploader to be found first.
Timeline of the AI Voice Cloning Lawsuit Against TikTok
The case has taken more than two years to reach judgment. Most of that time went on a failed attempt to unmask the uploader. The table sets out the key dates from Japanese press coverage of the complaint, the justice ministry and AFP.
| Date | Event |
|---|---|
| July 2024 | First of 188 narrated videos appears on the anonymous TikTok account |
| 15 October 2024 | 26 voice actors launch the NOMORE campaign against unauthorised generative AI |
| 1 April 2025 | Japan’s Information Distribution Platform Act takes effect; TikTok is designated as a large platform that month |
| June 2025 | Tsuda asks the Tokyo District Court to make TikTok disclose the uploader’s details |
| August 2025 | Court orders disclosure; TikTok hands over February 2025 connection records, but the internet provider’s logs have expired |
| September 2025 | Last of the 188 videos is posted |
| November 2025 | Tsuda sues TikTok’s operator for deletion of the videos |
| 24 April 2026 | Justice ministry study group on portraits and voices holds its first meeting |
| 23 May 2026 | The lawsuit becomes public in Japanese media |
| 7 August 2026 | Justice ministry publishes its final guideline on voices and generative AI |
| 30 September 2026 | Tokyo District Court verdict due |
Why the disclosure route failed
Japanese law lets a person whose rights are infringed online ask a court to order a platform to reveal who posted the material. Tsuda’s team started there in June 2025. In August 2025 the court accepted that his rights had been infringed for the purposes of that application and ordered disclosure.
TikTok then handed over connection records from February 2025. When Tsuda’s lawyers took those records to the internet provider, its log-retention period had already passed. The trail went cold, so in November 2025 Tsuda turned to TikTok itself and sued for deletion. This sequence is common in AI voice cloning disputes: the uploader is anonymous and the logs are short-lived, so the platform ends up as the only defendant within reach.
How long each stage took
The chart counts months from July 2024, when the first AI voice cloning video went up, to each milestone. It uses only the dates in the timeline above.
The account kept posting for another three months after Tsuda first went to court. For performers, the gap between a clone going live and any legal remedy is the part of AI voice cloning that hurts most.
The Money Behind the AI Voice Cloning Account
Tsuda’s lawyers argue that the account made money from his voice rather than from its content. According to court records seen by AFP, they said its posts “mostly lacked original substance” and “only gained popularity because its dubious, sordid content was narrated by the attractive voice of a popular actor”.
The account by the numbers
The figures below come from the complaint as reported by AFP and the Japanese press. The total view count is our own arithmetic, multiplying the reported video count by the reported average.
| Measure | Figure | Source |
|---|---|---|
| Videos posted | 188 | Complaint, via Japanese press |
| Posting period | July 2024 to September 2025 (15 months) | Complaint, via Japanese press |
| Followers at peak | More than 200,000 | Complaint, via AFP |
| Average views per video | About 1.47 million | Nippon TV report |
| Implied total views | About 276 million (188 × 1.47 million) | Our arithmetic |
| Estimated monthly revenue | More than ¥500,000 (about $3,200) | Complaint, via AFP |
Some Japanese outlets put the monthly income at ¥500,000 to ¥750,000. AFP’s figure is the floor, so we use it below.
What the complaint’s revenue figure adds up to
If the account earned the complaint’s floor figure every month, this is how the total would build over the 15-month posting period. It is illustrative arithmetic, not a finding. The account may have earned less in its early months, and the real total may be higher or lower.
Why cloned narration pays
The business model behind this kind of AI voice cloning is not new. Faceless channels on TikTok and YouTube have long paired stock images with a narrator. What has changed is the cost of a famous-sounding narrator, which is now close to zero. A recognisable voice makes viewers stop scrolling, and on a platform that pays by views, stopping is what earns money.
That is also why the court’s view on “customer attraction” matters so much. If the voice is the reason people watched, the case for a publicity-rights breach is much stronger. If people watched for the urban legends, TikTok’s defence gets easier.
How TikTok Is Defending Itself Against the AI Voice Cloning Claim
TikTok has not argued that AI voice cloning is harmless. It has argued that this particular voice is not Tsuda’s, that he is not famous enough for one of the laws he relies on, and that a platform should not be held responsible for what its users post.
“A generic male voice”
TikTok’s first line of defence is similarity. It argues that the narration is a “generic male voice” whose perceived similarity to Tsuda is subjective at best. It also says many people have a similar voice, and that viewers commented on the content of the videos rather than the voice.
The anonymous poster has added a twist. On an external site, according to Nikkei, the uploader said the narration was made by training an AI model on a friend’s voice. If true, the system never heard Tsuda at all. The court still has to decide whether that matters when the result sounds like him.
Questioning Tsuda’s “national fame”
TikTok also rejected the claim that the account was free-riding on a well-known brand, in breach of Japan’s Unfair Competition Prevention Act. It questioned whether Tsuda has “national fame”. That phrase matters because the Act has two relevant clauses with different thresholds, covered in the legal section below.
“A forum” for free speech
Finally, TikTok told the court that, as a platform, all it did was “host a forum where users could post creative and ingenious content within the scope of free speech”. This is the classic intermediary argument: the platform did not make the videos and should not be treated as if it did.
How the arguments line up
| Issue | Tsuda’s position | TikTok’s position |
|---|---|---|
| Is it his voice? | Acoustic analysis shows high similarity; viewers named him in comments | A generic male voice; the uploader says it was trained on a friend |
| Did the voice draw viewers? | The content lacked substance, so the voice drove popularity | Viewers were drawn by the topics, not the narrator |
| Unfair competition | The account free-rode on a well-known performer’s identity | Tsuda lacks the national fame the law requires |
| Platform duty | TikTok should delete infringing posts | It only hosts a forum for free expression |
Japan's Legal Tests for AI Voice Cloning
Japan has no statute that names the human voice as protected property. Tsuda’s lawyers are therefore relying on two existing routes: publicity rights, which Japan’s courts built up case by case, and the Unfair Competition Prevention Act. Copyright offers little help, for reasons explained below.
Publicity rights and the Pink Lady ruling
The leading case is the Supreme Court’s Pink Lady judgment of 2 February 2012. The court held that a person’s name and portrait are “the symbol of one’s personality” and that the right to exploit their customer-attracting power belongs to that person. It named three typical infringements: using a likeness as a product in itself, attaching it to goods to set them apart, and using it in advertising.
The question for any AI voice cloning claim is whether “portrait, etc.” covers a voice, and how a court tells a copied voice from a similar one. A Cabinet Office intellectual property panel said in 2024 that voices can be protected under publicity rights. Until now, no court had applied that view to AI voice cloning.
The Unfair Competition Prevention Act
The Act bans two relevant practices. Article 2(1)(i) covers using an indication that is well known among consumers in a way that causes confusion. Article 2(1)(ii) covers misusing an indication that is famous, which courts read as recognised nationwide, whether or not anyone is confused.
TikTok’s challenge to Tsuda’s “national fame” is aimed at the higher bar. According to Japanese reports, the Ministry of Economy, Trade and Industry said in 2025 that the Act could be used against AI that reproduces a person’s voice, for example to make it sing songs the person never recorded.
Why copyright does not help
Japan’s Copyright Act protects works and recorded performances, not the quality of someone’s voice. AI voice cloning from scratch, or from a friend’s recordings as the uploader claims, copies no recording by Tsuda. That leaves the personality-based routes as the only realistic path for AI voice cloning claims like this one.
The three routes compared
| Legal route | Source | What must be shown | Fit with this case |
|---|---|---|---|
| Publicity rights | Case law (Pink Lady, 2012) | An identifiable likeness used mainly for its customer-attracting power | Strong, if the court accepts the voice is identifiably his |
| Unfair competition, well-known indication | Article 2(1)(i) | A well-known indication used in a way that causes confusion | Possible; depends on the voice working as an indication |
| Unfair competition, famous indication | Article 2(1)(ii) | Nationwide fame; no confusion needed | Contested; TikTok disputes his national fame |
| Copyright | Copyright Act | Copying of a protected work or recording | Weak; no recording of his was copied |
What the Justice Ministry's AI Voice Cloning Guideline Says
On 7 August 2026, Japan’s Ministry of Justice published a 143-page report from its study group on civil liability for unauthorised use of portraits and voices. It is subtitled an interpretive guideline on publicity-rights infringement by generative AI. It is not a new law and carries no penalties. But it is the most detailed official statement yet on how existing rights apply to AI voice cloning.
How the guideline was made
The ministry announced the study group in April 2026 and held the first meeting on 24 April. The group met five times, published a draft on 27 July and issued the final report on 7 August. Meiji University’s Tetsuya Imamura, a member, told the Yomiuri Shimbun in May that similarity is easy to judge for signatures and faces but hard for voices. He said court rulings in concrete cases would be needed to move voice protection forward.
Voice as “the symbol of one’s personality”
The guideline states that a voice, like a portrait, identifies a person and works as a symbol of their personality. It is therefore covered by publicity rights and by the personality-based right not to have one’s likeness used without good reason. AFP quoted the ministry saying that as “the symbol of one’s personality”, a voice is subject to “publicity rights” and “rights to prevent unauthorised use of one’s portrait and likeness”.
Similarity plus context decides identification
Because AI output is rarely identical to a real voice, the guideline says courts should look at how similar it is together with anything that points listeners to the person. That includes titles, tags and other context. The test is whether an audience can identify the voice as that person’s. In Tsuda’s case, the Jujutsu Kaisen-style profile picture and the comments naming him are exactly this kind of context.
Impressionists are safe; anonymous clones may not be
The report says impressions and voice mimicry generally do not infringe publicity rights. Impressionists usually perform under their own names, so the audience knows who they are watching. It adds a caveat: if someone uses a sound-alike without identifying themselves and audiences mistake it for the real person, there may be infringement. An unlabelled AI voice cloning account fits that caveat far better than a stage impressionist does.
A voice actor does not need a famous face
One footnote goes straight to TikTok’s fame argument. If a voice actor’s voice itself drives sales, the guideline says, it has customer-attracting power. It is not necessary for the actor’s name or face to be widely known. The footnote is about publicity rights, not unfair competition, but it undercuts the idea that a voice actor must be a household name before AI voice cloning can infringe his rights.
Platforms can be asked to delete
The report says requests to delete content that infringes publicity rights can also be made to the platform operator that manages it. It also treats earning advertising revenue from social media videos as commercial use. A site that makes money from many videos of a celebrity, it says, can fall within the Pink Lady categories.
What the guideline means for this case
| Guideline point | Relevance to Tsuda v TikTok |
|---|---|
| Voice is covered by publicity rights | Removes the threshold question of whether a voice can be protected at all |
| Identification uses similarity plus context | The character-style avatar and viewer comments support identification |
| Impressions are generally lawful | Does not help an anonymous account that never named its narrator |
| A voice actor’s face need not be famous | Weakens the argument that Tsuda is not famous enough |
| Deletion can be sought from the platform | Supports suing TikTok when the uploader cannot be found |
| Advertising revenue counts as commercial use | Fits an account said to earn over ¥500,000 a month |
The guideline is non-binding. As AFP noted, it remains to be seen whether and how it influences Wednesday’s decision. Judges can depart from it, and much of the case was argued before it was published.
How the Tokyo Court Could Rule on AI Voice Cloning
There are three broad ways the verdict could go. Each would send a different signal about AI voice cloning to performers, platforms and the people who run faceless channels.
Outcome 1: publicity right recognised, deletion ordered
The court could accept that Tsuda’s voice is protected, that the narration identifiably copies it, and that TikTok must delete the videos. That would be the first judicial recognition in Japan that AI voice cloning of a performer can breach publicity rights. Platforms would face deletion claims without the uploader being identified first.
Outcome 2: right recognised, similarity not proven
The court could agree that a voice can be protected but find that this narration is not close enough, or that viewers came for the content. Performers would win the principle and lose the case. Future plaintiffs would learn how much acoustic and contextual evidence they need.
Outcome 3: claim rejected
The court could hold that TikTok was not obliged to delete, or decline to extend publicity rights to a synthetic voice. Voice actor Bin Shimada warned AFP that “our industry can be crushed” if a win for TikTok triggers “unbridled” AI voice cloning on social media. Professor Ueno thinks outright denial is “very unlikely”.
| Outcome | For performers | For platforms | For AI voice cloning channels |
|---|---|---|---|
| Deletion ordered | A usable precedent and a direct route to removal | Must act on credible voice complaints | Sound-alike narration becomes a legal risk |
| Right recognised, claim fails on facts | Principle won; evidence bar set high | Can defend on similarity grounds | Uncertain; “a friend’s voice” defences multiply |
| Claim rejected | Must wait for legislation or appeal | Hosting defence strengthened | Little legal deterrent in Japan |
Why Voice Actors Call AI Voice Cloning an Existential Threat
Japan’s voice actors, known as seiyu, have campaigned on this issue for two years. For them, AI voice cloning is not an abstract intellectual property question. It is a direct threat to how they earn a living and to their control over what their voice appears to say.
The NOMORE campaign
On 15 October 2024, 26 voice actors launched a campaign called “NOMORE Mudan Seisei AI”, or No More Unauthorised Generative AI. Participants included Koichi Yamadera, Yuki Kaji and Jun Fukuyama. They called for rules on unauthorised commercial use of AI-generated voices and video, and said they wanted to find a way to coexist with AI.
Bin Shimada, who is part of the group, told AFP that unchecked AI voice cloning could cause younger actors to “lose hope”. Fukuyama told the Yomiuri that he had seen videos in which characters he voiced were made to sing without permission. He said it gave him “the terror of not being able to control my own voice”.
The Japan Actors Union
Yuko Sasaki, a voice actor and executive director of the Japan Actors Union, told AFP that an actor’s voice “is the result of years of rigorous training and apprenticeship”. “It’s something we spend lots of time and money perfecting,” she said, adding that “we support Tsuda wholeheartedly”. The court, she said, should “acknowledge that ownership of a voice is a fundamental right”.
When the draft guideline came out in July, the union welcomed it and set out five demands for the final version. They were prior, specific consent for AI use of a likeness or voice; clear duties for AI providers and platforms; fast remedies including deletion and blocking re-uploads; protection regardless of fame or profit motive; and protection for the dead.
“That’s our only weapon”
Michihiro Ikemizu, an 83-year-old veteran voice actor, expressed solidarity with Tsuda but struck a different note. “You need to constantly feel and adapt while performing your role. That’s something AI can never do,” he told AFP. “That’s our only weapon.”
The counter-argument: who owns a type of voice?
Not everyone thinks broad voice rights are simple. Writing for Yahoo! News Japan, tech journalist Shuji Shinohara said some protection should exist but warned of side effects. Many voice actors sound alike. A wide right could restrict younger performers or streamers who happen to sound like a star. Badly designed, it could even enable “voice patent trolls” who claim a type of voice and demand fees. Any AI voice cloning rule has to protect identity without giving anyone a monopoly on a timbre.
How Other Countries Treat AI Voice Cloning
Japan is not alone. Courts and legislators in several countries have already had to decide whether a voice belongs to the person who speaks with it. Performers’ groups argue that copyright and data protection rules have not kept pace. The International Federation of Actors said last year that “in many countries copyright and data protection laws are not being adequately adapted” to protect performers from AI voice cloning.
| Jurisdiction | Case or law | Date | What it covers | Status |
|---|---|---|---|---|
| Japan | Justice ministry guideline | 7 Aug 2026 | Voice under publicity and personality rights | Non-binding; first ruling due 30 Sep |
| China | Beijing Internet Court, voice artist Yin | 23 Apr 2024 | Personality rights extend to identifiable AI voices | 250,000 yuan awarded |
| India | Bombay High Court, Arijit Singh | 26 Jul 2024 | Name, voice and vocal style against AI voice models | Interim injunction |
| US, New York | Lehrman and Sage v Lovo | 10 Jul 2025 | State publicity and consumer-protection claims over voice clones | Proceeding; federal trademark claims dismissed |
| US, Tennessee | ELVIS Act | In force 1 Jul 2024 | Adds voice to the state publicity right | In force |
| US, federal | NO FAKES Act of 2026 | 18 Jun 2026 | Federal right over digital replicas of voice and likeness | Cleared Senate committee; awaiting floor vote |
| US, California | SB 1050 | Signed 16 Sep 2026 | Ad disclosure for synthetic performers, including voices | Operative 1 Jan 2027 |
| Denmark | Copyright Act amendment | Proposed Jun 2025 | Body, face and voice, for 50 years after death | Expected July 2026, per EPRS |
| European Union | AI Act Article 50 | 2 Aug 2026 | Disclosure of deepfake audio and video | Applies |
| United Kingdom | Passing off (Irvine v Talksport) | 2002 | False endorsement only | No standalone personality right |
China: the first AI voice judgment
On 23 April 2024, the Beijing Internet Court ruled in China’s first case on AI-generated voices. A voice artist surnamed Yin found her voice in audiobooks after recordings she made were used to build a text-to-speech product. The court held that protection of a natural person’s voice extends to AI-generated voices if the public can identify them as that person. Two defendants were ordered to pay her 250,000 yuan.
India: an injunction for a singer
On 26 July 2024, the Bombay High Court granted singer Arijit Singh an interim injunction. It barred unauthorised use of his name, voice, vocal style and likeness, including through AI voice models and voice conversion tools. It was India’s first AI voice cloning personality-rights ruling.
United States: patchwork, then maybe a federal right
In New York, voice actors Paul Lehrman and Linnea Sage sued the AI voice firm Lovo. In July 2025 a federal judge dismissed their federal trademark claims and most copyright claims. Their state right-of-publicity and consumer-protection claims were allowed to proceed. Tennessee’s ELVIS Act, in force since July 2024, adds voice to the state’s publicity right.
At federal level, the NO FAKES Act of 2026 cleared the Senate Judiciary Committee by unanimous voice vote on 18 June 2026. It would create a federal right over digital replicas of a person’s voice and likeness. California has taken a disclosure route instead. Its SB 1050 requires ads to label synthetic performers, including voices, from 2027, as covered in our article on the California synthetic performer law.
Europe: copyright for your face and voice
Denmark proposed in June 2025 to amend its Copyright Act so that every person has rights over their body, facial features and voice. The European Parliament’s research service said in January 2026 that the change was expected to take effect in July 2026, with protection lasting 50 years after death. At EU level, Article 50 of the AI Act has required disclosure of deepfake audio and video since 2 August 2026. Our guide to EU AI content label rules explains how those duties work.
The Scarlett Johansson precedent
The best-known dispute did not reach court. In 2024 OpenAI apologised to Scarlett Johansson over “Sky”, a ChatGPT voice that sounded like hers, released after she had declined to work with the company. OpenAI paused the voice. The episode showed that the sound-alike problem is the core AI voice cloning risk even when no recording is copied.
TikTok's Own Rules on AI Voice Cloning
TikTok’s community guidelines already address AI voice cloning, and their wording lines up closely with the arguments in court. The current version defines “likeness” as a recognisable image, video or audio representation of a person, “including their face, body, voice, and gestures”.
Disclosure when audio mimics a real person
The guidelines require creators to label AI-generated or significantly edited content that shows realistic people or scenes. Disclosure is needed when “AI-generated audio mimics the voice of a real person”. Unlabelled content “may be removed, restricted, or labeled” by TikTok’s team, depending on the harm it could cause.
The generic text-to-speech exemption
The same section says disclosure is not needed when “using generic text-to-speech (TTS) narration, when the TTS isn’t a recognizable voice of a known individual”. Text-to-speech engines grew out of natural language processing research, and generic narration voices are everywhere on the platform. TikTok’s defence in court, that the narration is a “generic male voice”, is the same test as its own policy, applied to the same facts.
That makes the ruling doubly important for the platform. If the court finds the voice recognisably Tsuda’s, it will in effect say that TikTok’s own policy required a label that was never there.
Japan’s platform law
Japan’s Information Distribution Platform Act took effect on 1 April 2025. That month the government designated TikTok Pte. Ltd., along with Google, LINE Yahoo, Meta and X, as a large-scale platform operator. Designated operators must publish how victims can request deletion of infringing posts and must, in principle, tell the requester what they have decided within seven days. The Act does not decide whether AI voice cloning infringes anyone’s rights, but it gives performers a faster channel once a court or guideline says it does.
What the AI Voice Cloning Case Means for Creators, Platforms and Brands
Whatever the verdict, the Tsuda case has already changed the risk picture for anyone who makes, hosts or buys synthetic voices. The guideline sets out the reasoning; the ruling will show how far a court will take it.
For voice actors and agencies
Performers should document their voice in a form a court can use, such as clean reference recordings and existing acoustic profiles. Contracts should spell out whether recordings may be used to train AI, for what and for how long. The Japan Actors Union now runs a legal contact point for reports of unauthorised use, and says it is receiving a very large number of them.
For platforms
A platform that relies on uploader self-labelling will struggle if courts treat a recognisable voice as identifiable. Voice-similarity screening against a register of consenting performers, faster action on credible complaints and consistent enforcement of existing labelling rules would all reduce exposure to AI voice cloning claims.
For brands and businesses using synthetic voices
Companies that use generated narration in ads, apps or support lines should use licensed voices with documented consent. They should avoid briefs that ask for a voice “like” a named actor, and label synthetic audio where the law or platform rules require it. A voice that is legal in one market may not be in another. Teams building an AI strategy should treat voice rights as a procurement check, not an afterthought.
Cloned voices are also a fraud tool. Our report on the cloned-voice messaging scam that cost Intesa millions shows how the same technology is used against businesses.
| Who | Main AI voice cloning risk | Practical step |
|---|---|---|
| Voice actors | Anonymous clones earning from their voice | Keep reference recordings; set AI terms in every contract |
| Talent agencies | Loss of licensing value | Monitor platforms; use deletion channels quickly |
| Platforms | Deletion orders and liability claims | Enforce voice labels; screen for recognisable voices |
| Creators | Takedowns and damages claims | Use licensed or clearly generic voices; label synthetic audio |
| Brands | Sound-alike ads and disclosure breaches | Document consent; avoid “sounds like” briefs |
What Comes After the AI Voice Cloning Verdict
Wednesday’s judgment is a first-instance decision. Under Japan’s Code of Civil Procedure, either side can appeal within two weeks of receiving the written judgment, so the final word may be some way off.
What to watch on 30 September
Three points in the judgment will matter most. First, whether the court accepts that a voice falls within publicity rights, and how it describes the identification test. Second, whether it treats TikTok as obliged to delete once it knows of infringing posts. Third, whether it addresses the Unfair Competition Prevention Act and the “national fame” question, or decides the case without it.
The global signal
Anime is one of Japan’s biggest cultural exports, and its voice actors have large followings abroad. A clear ruling would give performers elsewhere a reference point. Courts in China, India and the United States have already moved. Europe is legislating on labels and likeness. Japan now has a guideline and, within days, its first judgment on AI voice cloning.
For businesses, the direction of travel is the same everywhere: voice is becoming a protected attribute. Our overview of deepfake defences covers the detection side of the problem.
AI Voice Cloning and the Tsuda Case: Frequently Asked Questions
Who is suing TikTok over AI voice cloning?
Kenjiro Tsuda, a 55-year-old Japanese voice actor known for Jujutsu Kaisen and Yu-Gi-Oh!. He sued TikTok’s operator in the Tokyo District Court in November 2025, asking for deletion of videos he says used AI voice cloning to copy his voice.
When is the verdict?
The Tokyo District Court is due to rule on Wednesday 30 September 2026.
Is AI voice cloning illegal in Japan?
No law bans it outright. The justice ministry’s August 2026 guideline says unauthorised use of an identifiable voice can infringe publicity rights and personality rights, especially when it is used commercially. The Tsuda ruling will be the first court test of that view.
What does TikTok argue?
TikTok says the narration is a “generic male voice”, questions whether Tsuda has the national fame the unfair competition law requires, and says it only hosts a forum for users’ free expression.
Do voice impressions count as AI voice cloning?
Not in the guideline’s view. Impressionists usually perform under their own names, so audiences know the voice is not the original. The AI voice cloning risk arises when a sound-alike is used without that disclosure and audiences mistake it for the real person.
References and Further Reading
Japanese anime voice actor takes TikTok to court over AI voice clone (Kuwait Times, AFP)
Japanese anime actor fights TikTok over AI voice cloning (Aaj English TV, AFP)
Tokyo District Court to rule on 30 September in generative AI voice case (Cedar News, Japanese)
Kenjiro Tsuda sues over AI voice imitation: similarity and rights at issue (audiostart, Japanese)
The side effects of legally recognising voice rights (Shuji Shinohara, Yahoo! News Japan)
Justice ministry states unauthorised use of voices infringes rights (ITmedia AI+, Japanese)
Japan Actors Union statement on the justice ministry guideline (Nippairen, Japanese)
NOMORE Mudan Seisei AI campaign (official site, Japanese)
Voice actors launch NOMORE campaign video and website (ITmedia AI+, Japanese)
Integrity and Authenticity: Edited Media and AI-Generated Content (TikTok Community Guidelines)
Information Distribution Platform Act (Ministry of Internal Affairs and Communications, Japanese)
AI-generated voice mimicry infringes on rights (Beijing Internet Court)
New York court tackles the legality of AI voice cloning (Skadden)
Senate committee advances bill to protect name, image, likeness and voice (Holland and Knight)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
