Your Voice Is Now Your Likeness: Japan’s Landmark AI Voice-Cloning Ruling and 8 Rules Before Your Business Uses a Synthetic Voice (AI Trends, 30 September 2026)

What’s trending in AI on 30 September 2026: a court has said out loud what performers have argued for years. The Tokyo District Court ruled today that a human voice is protected under Japan’s publicity rights, the same doctrine that protects a person’s name and face. The case was brought by Kenjiro Tsuda, the voice actor behind characters in Jujutsu Kaisen and Yu-Gi-Oh!, against TikTok, over an anonymous account that posted 188 videos narrated by an AI voice that sounded like his. It is the first ruling of its kind in Japan. And yet Tsuda’s claim was dismissed, because the account had already disappeared. This post explains what the court decided, why a “loss” is still a landmark, how voice rights now compare in Japan, the US and the EU, and eight rules for any business that uses, or might be imitated by, a synthetic voice.

Key takeaways

Voice joins name and face. The court said a voice symbolizes a person’s identity much as a portrait does, and that using it without permission to exploit its commercial pull can infringe publicity rights.
A principle, not a payout. The takedown request was dismissed as moot because the account was deleted mid-case, and no damages were awarded.
The test is recognition, not voiceprints. Under Japan’s August 2026 guidance, what matters is whether listeners recognize whose voice it is. Forensic analysis helps but isn’t decisive.
Businesses are exposed on both sides. Using a sound-alike voice in marketing is now a legal risk in more markets, and your own executives’ voices need protecting too.

Japan now treats a recognizable voice like a face or a name when it’s used to make money.

What happened in the Tsuda v. TikTok case

Tsuda, 55, is known in Japan for a deep, distinctive voice. According to AFP and Inven Global, an anonymous TikTok account used generative AI to narrate videos about urban legends and conspiracy theories in a voice that closely resembled his. The account posted 188 videos, built more than 210,000 followers and averaged about 1.47 million views per video. Tsuda’s legal team estimated it earned ¥500,000 to ¥750,000 a month, roughly $3,200 to $4,800.

Because the account holder was anonymous, Tsuda sued the platform, asking for the videos to be removed and arguing that they infringed his publicity rights and breached Japan’s Unfair Competition Prevention Act. His lawyers even used deep-learning models to compare the narration with his real voice. TikTok countered that the narration was a generic male voice and any resemblance was subjective.

A sound-alike voice built a sizeable, paying audience before anyone could stop it.

What the court decided, and why Tsuda still “lost”

Presiding Judge Aya Takahashi held that a person’s voice is “as symbolic of one’s personality as portraits are,” according to AFP. Using a performer’s voice without permission, where the purpose is to exploit its commercial appeal, can therefore infringe publicity rights. That wording mirrors the test Japan’s Supreme Court set for celebrity photographs in its 2012 Pink Lady decision, now extended from faces to voices.

The claim itself was dismissed. The uploader deleted the account while the case was running, so there was nothing left to take down, and the court did not decide whether Tsuda’s own rights had actually been infringed. No damages were awarded. His lawyer told Inven Global the original goal, getting the videos gone, had been achieved, and that the ruling should help future plaintiffs. The Japan Actors Union, which backed the suit, stressed that a professional voice is the product of years of training.

The precedent matters more than the outcome: the next plaintiff starts with voice already recognized as protected.

The test that matters: would listeners recognize the voice?

The ruling lands on top of non-binding guidance Japan’s Justice Ministry published on 7 August 2026 after a study group convened in April. According to FourWeekMBA’s summary, the guidance treats voice as part of a person’s identity and lists the uses most likely to cross the line: narration presented as someone’s own voice, a voice used to set a product apart, and voices in advertising. Ordinary impressions are fine when the impersonator is named and it’s their skill that draws the audience.

Crucially, the test is whether the audience recognizes the voice as belonging to a specific person, not whether a lab can match its voiceprint. Forensic analysis may count as one factor, but it isn’t decisive. For businesses, that means “we trained our own model” is no defense if customers hear a famous voice.

Where voice rights stand in Japan, the US and the EU

Three different routes to the same destination: a recognizable voice is no longer free raw material.

United States. There is still no federal voice right. States fill the gap, with Tennessee’s ELVIS Act the best-known law aimed at AI voice clones. The federal NO FAKES Act would give everyone, not just celebrities, control over digital replicas of their voice and likeness, with a DMCA-style notice-and-takedown process and liability for platforms that knowingly host unauthorized replicas. Holland & Knight reports the Senate Judiciary Committee advanced it unanimously on 18 June 2026; it still needs a full Senate vote and the House.
European Union. The EU AI Act’s transparency obligations, which apply from August 2026, require AI-generated or manipulated content that imitates real people, including audio deepfakes, to be disclosed as such. Personality and data-protection laws in member states add further protection.
Japan. No statute yet, but now guidance plus a court ruling pointing the same way, which is often how Japanese law evolves.

This is general information, not legal advice. If you’re planning a campaign or product that uses synthetic voices, have counsel review it for each market you sell in.

What this means for your business

Marketing teams: “sounds like” is now a risk category. Asking an AI voice tool for a voice similar to a well-known actor or influencer is exactly the use the Japanese test targets. The same logic runs through US state laws.
Content creators and agencies: platforms will act faster. The Tsuda account disappeared once a lawsuit arrived. Expect TikTok, YouTube and others to remove sound-alike content more readily as voice rights firm up, taking your audience with it.
Customer service: disclose your AI voices. After last week’s wave of voice AI launches, many businesses are putting synthetic voices on phone lines. Disclosure is now required in some markets and is simply good practice everywhere.
Leaders: your own voice is an asset to protect. The same technology that cloned Tsuda can clone your CEO for a payment-fraud call. Voice rights help after the fact; verification procedures stop the money leaving in the first place.

8 rules before your business uses a synthetic voice

Never brief for a sound-alike. Ban prompts, scripts and creative briefs that ask for a voice “like” a named person. Record that rule in your brand and AI-use policy.
Get specific, written consent for any real voice you clone. Include the channels, markets, duration, payment and how the person can withdraw consent. That applies to employees and executives too.
Check where your vendor’s voices came from. Ask AI voice providers whether stock voices were recorded with licensed consent, and whether they indemnify you if a voice turns out to resemble someone.
Run a recognition test. Before launch, play the voice to a few people outside the project. If anyone names a real person, change the voice. That mirrors the test Japan’s guidance uses.
Disclose AI voices. Label AI narration in ads and videos, and tell callers when they’re talking to an AI agent. It’s required under EU rules and expected by customers everywhere.
Keep the paperwork. Store consent forms, vendor licenses, voice model versions and the prompts used for each published asset, so you can answer a complaint in hours, not weeks.
Protect your people’s voices. Limit long public recordings of executives where you can, add call-back or code-word checks for payment and access requests, and train finance and help-desk staff on voice-clone fraud.
Have a takedown playbook. Know each platform’s impersonation and IP reporting routes, who in your company files them, and when to escalate to counsel. If the NO FAKES Act passes, a formal notice-and-takedown process will follow.

For the wider picture on generative content and rights, see our guide to generative AI in content creation and how AI is changing social media. And if your AI vendors are among those that signed today’s White House AI Accord, add a question about voice cloning safeguards to your vendor review.

Frequently asked questions

What did the Tokyo court rule about AI voice cloning?

On 30 September 2026, the Tokyo District Court ruled that a human voice is protected under Japan’s publicity rights, like a person’s name or face, and that using someone’s voice without permission to exploit its commercial appeal can infringe those rights. It was the first ruling of its kind in Japan.

Why was Kenjiro Tsuda’s case dismissed if voice is protected?

Tsuda asked for the AI-voiced videos to be removed, but the anonymous uploader deleted the account while the case was underway. With nothing left to remove, the court dismissed the claim without deciding whether his own rights had been infringed. No damages were awarded.

Is it legal to use an AI voice that sounds like a celebrity?

It is increasingly risky. In Japan, a recognizable voice used for commercial appeal can now infringe publicity rights. In the US, state laws such as Tennessee’s ELVIS Act protect voice, and the proposed NO FAKES Act would add federal rights. In the EU, AI-generated audio imitating real people must be disclosed. Get legal advice for your markets before using any sound-alike.

What is the NO FAKES Act?

The NO FAKES Act is a proposed US federal law that would give every person rights over AI-generated replicas of their voice and visual likeness, with a notice-and-takedown process modeled on copyright law and exceptions for news, parody and criticism. The Senate Judiciary Committee advanced it in June 2026; it has not yet become law.

How can a business use AI voices safely?

Use licensed or consented voices only, never brief for a sound-alike, test whether listeners recognize the voice as a real person, disclose AI voices to customers, keep consent and license records, and have a takedown plan in case your own people’s voices are cloned.

Delana Technologies