AI law and the newsroom, September 29–30: a Tokyo court puts voice under publicity rights, Brussels reopens the copyright file, and California's last hours

This issue covers 29–30 September 2026: four items from three jurisdictions. A Tokyo court held that a person's voice is protected in Japan. The European Commission reopened the copyright file that governs AI training. The trade body for US news publishers went public in its fight with the government. And California's governor reached the last hours of his deadline for a stack of AI bills.

Three questions run through the items below, and this digest keeps asking them. What may AI companies do with your work? What must you tell your audience when you use AI? Which office, and which document, will ask you about it?

The Tokyo District Court ruled on 30 September that a person's voice falls under publicity rights — the right to control the commercial value of one's own name and likeness — and that unauthorised use of a performer's voice can infringe them. "A person's voice, similar to a likeness, is a symbol of the individual's personality," presiding judge Aya Takahashi wrote. A Japanese court has said so for the first time. 1

The same judgment dismissed the claim in front of it. Voice actor Kenjiro Tsuda had asked the court to order TikTok to delete videos narrated by an imitation of his "deep and lustrous" voice. Judge Takahashi found the platform was "no longer obliged to delete" them, because the anonymous account behind them had taken them down several months after Tsuda sued in November 2025. 2 The court settled the right and stopped short of any finding on Tsuda's own case. 1

The complaint described 188 videos posted from a single account between July 2024 and September 2025, an audience of about 210,000 subscribers, and an average of roughly 1.47 million views per video, enough to earn the operator between ¥500,000 and ¥750,000 a month. 3 Tsuda's lawyer, Takashi Hirano, told reporters the ruling "ended up making both sides look victorious". TikTok said it would respect the judgment. 2

Japan had already moved this way. In August the Justice Ministry's expert panel issued guidance saying voices should be protected under publicity rights in the same way as images, and that uploading fake videos imitating an actor's or singer's voice in order to earn money can be a rights violation. 4 That guidance is advisory. The ruling binds the two parties to this case, and its finding on the right is the part with reach beyond them.

Judge Takahashi framed the protection around purpose: it applies where the point of using a voice is to take advantage of its commercial appeal. 2 For a newsroom, that framing is the part to hold. A dubbing track, an AI-read audio article modelled on a presenter, or a reconstructed interview all raise the claim Tsuda brought. Since 30 September, a newsroom using a synthetic voice for commercial effect in Japan faces a court holding that the right exists, and a ministry document from August that points the same way. 1

On 29 September the European Commission opened a targeted consultation on whether the EU's copyright rules need reinforcing as technology, artificial intelligence included, changes the market. Its first subject is the use of copyright-protected content in AI. The others are piracy of live events such as sport, the single equitable remuneration right of music performers and producers, and copyright in scientific research. 5

Rights holders, generative AI providers, intermediaries, collective management organisations, research organisations, national authorities, consumer organisations and non-governmental organisations are all invited to answer. Responses close on 3 November 2026. 5

The consultation gathers evidence, and any measure would come after it. The Commission says it continues an earlier call for evidence, and asks whether the framework should be reinforced through additional measures. 5 For a publisher, this is the place to say what the rules on training data should be. The window for saying it closes on 3 November.

Two days after The New York Times and the other news plaintiffs answered the Justice Department in court, their trade body endorsed the answer in public. On 29 September the News/Media Alliance, which represents US news publishers, issued a statement backing the 28 September filing. That filing asks Judge Sidney Stein to give the department's statement of interest — an agency's view filed in a case it is not party to — "no weight at all". 6

The alliance calls the department's submission contradicted by the evidence in the case and inconsistent with the Copyright Office's own analysis. Its president, Danielle Coffey, argued that a government which expects data centres to pay for electricity and chipmakers to be paid for their products should expect AI companies to pay for news. 6 The department filed the statement on 1 September, urging the court to find that training AI models on copyrighted works is fair use. 7

Both documents are argument, and the ruling that will carry weight for a newsroom is still to come. Judge Stein holds both sides' summary-judgment motions in the consolidated OpenAI copyright case, and his decision will settle how much the government's reading of fair use counts. 7

California's governor has until midnight in Sacramento on 30 September to sign or veto the AI bills still on his desk. 8 Three of them bear on publishing. SB 1000 rewrites the California AI Transparency Act: it drops the one-million-monthly-visitor threshold from the definition of a covered provider, which widens the set of AI operators the Act reaches, and it removes the duty to offer visible disclosure while keeping the duty to embed machine-readable provenance data. AB 2713 would require large platforms to give users an interface showing whether content was generated or substantially altered by AI, and to let them download the underlying provenance data. AB 1609 would require businesses with more than $500 million in annual national revenue to disclose when a customer is dealing with a chatbot, and to offer a route to a human. 8

As this issue went out, the AI package remained on the governor's desk. Inaction has a known result in California: a bill the governor leaves unsigned becomes law without his signature, the law firm Venable notes in its September roundup of the package. 9 SB 1000 carries an urgency clause, so it would take effect on enactment; the rest of the package lands on 1 January 2027. 8

The nearest date to hold is 3 November 2026, when the Commission's copyright consultation closes. Then 2 December 2026, when further AI Act enforcement begins for the transparency obligations and the prohibited-practice rules. Then 1 January 2027, when the rest of California's AI package takes effect, if the governor lets it through.