On 30 September 2026 the Tokyo District Court (presiding judge Aya Takahashi) held that a person's voice can be protected under publicity rights. It described the voice as a "symbol of personality", comparable to a likeness, when it is used to exploit its commercial appeal. The plaintiff was voice actor Kenjiro Tsuda, known for voicing Kento Nanami in Jujutsu Kaisen. He did not win the remedy he asked for, and the court did not decide whether his own rights had been infringed. The reasoning is still important. Japan has no AI-specific liability statute, and this is the first judicial statement of how its civil law reaches synthetic voices.
What the court actually decided
According to The Next Web's account, an anonymous TikTok user posted 188 videos between July 2024 and September 2025. They paired images with AI-generated narration about urban legends and the occult, and the account peaked at over 200,000 followers. The profile picture resembled a Jujutsu Kaisen character Tsuda voices. TikTok argued that the narration used a "generic male voice" and that any resemblance to Tsuda was subjective.
The court rejected Tsuda's takedown request because the account had already been deleted in May 2026. It left the question of infringement of his own rights undecided, and he did not seek damages. His lawyer said the outcome made "both sides look victorious". This is a doctrinal ruling, not a finding that this particular clone was unlawful.
Why a narrow, existing-law approach is the right one
The strongest case for fast, bespoke legislation is easy to state. Voice clones are cheap and convincing. Performers face real income losses, and a case-by-case lawsuit takes years. By the time Tsuda's case reached judgment, the videos were already gone. A new statute with statutory damages and a notice-and-takedown duty could give faster relief.
The court's route shows why that may be unnecessary. Japan's publicity right was never codified. It comes from the Supreme Court's 2012 "Pink Lady" judgment, which WIPO Lex reproduces. That judgment found unauthorized use of a portrait tortious where its sole purpose is to exploit customer appeal, in three situations:
- use as a collectible product in its own right;
- use to differentiate merchandise;
- use to advertise goods or services.
In the Pink Lady case itself the photos only supported a magazine article on a weight-loss method, so the claim failed. The Tokyo ruling extends that same test to voice, and the extension keeps the test's limiting feature. Protection attaches to commercial exploitation of a persona. It does not attach to every resemblance, parody or imitation.
That limit matters for speech and innovation. A rule that protected a "voice style" against any similar-sounding output would chill impressionists, satirists, text-to-speech developers and the many creators who use synthetic narration for ordinary content. TikTok's "generic male voice" defence points at the hard evidentiary question of when a synthetic voice is similar enough to be treated as the performer's. A sole-purpose commercial test makes courts look at how a voice is used. That is a better question than whether a model could resemble someone.
The government is converging on the same view
The Justice Ministry reached a parallel conclusion. A study group set up in April 2026 met five times and published its report on 7 August 2026. It is titled as interpretive guidelines on publicity-right infringement involving generative AI. The group worked from current law and case law, and its hypothetical cases covered whether rights were infringed, the scope of damages, availability of injunctions and the Unfair Competition Prevention Act.
An earlier draft, reported in July, treated voice as protected like a likeness. Examples of infringement included AI audio mimicking singers or voice actors that is posted online to earn revenue or followers. Offering AI services that can generate voice actors' voices may infringe if those voices are used as a selling point in marketing. The same draft said human impersonation and voice mimicry are not infringement as long as the audience does not mistake the performer for the real person. METI's civil-liability guidance of 9 April 2026 is likewise non-binding interpretation of existing tort and product-liability rules. The court, the Justice Ministry and METI all arrive at interpretation, not a new regime.
What the ruling leaves open
The scale of the problem is real. The same report cites a nonprofit, the Japan Publicity Right Protection Organization, which counted more than 40,000 confirmed cases of unauthorized AI-linked images and videos of Japanese celebrities in a two-month period from late June 2025. It estimated losses of about ¥2 billion to ¥4.5 billion. A doctrinal ruling does not solve three practical gaps:
- Anonymity. The Tsuda defendant was an anonymous account. Suing a platform for removal works only while content is still up, and the content was deleted before judgment.
- Speed. Interim injunctions are the remedy that matters for viral clones, and the Tokyo ruling does not address them.
- Platform duties. The court did not say what a platform must do on receiving a credible voice-rights complaint.
These gaps are better closed by procedural fixes than by a new liability regime. Faster provisional-remedy practice and clearer platform notice procedures would help, as would courts or ministries publishing worked examples of the commercial-exploitation test. A broad new right over vocal likeness would invite overreach and litigation against legitimate synthetic-speech tools.
The takeaway
Japan is building AI voice-clone liability from an existing doctrine with a built-in commercial-purpose limit. The first appeal and the first damages award will show whether that limit holds under pressure. Until then, the sensible policy is to let courts work the test through, publish clear guidance, and resist legislating the whole problem before the evidence is in.