A Shanghai court ordered an unnamed generative AI developer to pay MiHoYo ¥750,000 (approximately $112,000) on June 30 after ruling the company had reproduced and sold the voices of 63 Genshin Impact characters without authorization. The defendant briefly appealed the order before withdrawing the challenge in the months since.

The verdict, first reported by Chinese outlet The Paper and relayed by Automaton Media, marks Shanghai's first decision on "unfair competition" arising from AI-recreated voiceover work. MiHoYo argued the defendant had bundled Genshin Impact "images and story animations" into commercial voice packs built on top of cloned performances.

What did the defendant actually sell?

MiHoYo's complaint detailed a generative AI service that replicated the vocal likenesses of 63 named characters drawn from Genshin Impact's roster. The defendant then distributed the cloned voices alongside the game's proprietary artwork and cutscene animation as paid downloads, according to the court.

The judge treated the bundle as coordinated commercial exploitation rather than incidental fan recreation. According to the court ruling reported by The Paper, the offering "demonstrated an intent to capitalize on the game's popularity."

How does this change the calculus for studios?

The judgment establishes precedent on several fronts for developers, publishers and tool vendors operating in China:

• Voice performances count as protectable unfair competition assets, not solely as copyrighted recordings.
• Bundling cloned voices with proprietary art, animation or narrative assets amplifies the infringement finding.
• The framework reaches third-party vendors that profit directly from cloned character material, even when individual modders go unscrutinized.

For studios shipping localized builds, the risk surface now extends well beyond internal voice recording. Chinese courts have shown willingness to treat synthesis plus commercial sale as a single infringing act.

Who else is exposed?

The same liability perimeter applies to third-party AI tooling pitched to the games sector. ElevenLabs, among other voice synthesis vendors, markets directly to studios for character dialogue, prototyping and accessibility workflows.

Voice actors Victoria Atkin and Tim Friedlander told Game Developer in 2023 that such tools had already enabled players to recreate their performances and ship mods inserting the cloned output into unrelated titles — sometimes for resale. The Shanghai verdict narrows the path for paid distribution that wraps AI clones in commercial packaging.

What remains unresolved?

The Shanghai judgment addresses only the sale and marketing of cloned characters. It does not reach the broader, still-pending question of whether copyrighted material ingested as AI training data constitutes fair use under Chinese, U.S. or EU law.

Courts in all three jurisdictions continue to weigh that question across separate proceedings. Chinese judges have yet to rule directly on training-data ingestion for synthetic voice models. Voice synthesis providers and studios face an unresolved liability line on this front.

For studios with Chinese operations or localization partners, MiHoYo's win gives legal counsel a workable enforcement template. Watch for the defendant's compliance deadline, the identification of any co-defendants, and follow-on civil suits MiHoYo files against resellers of similar packs.