01
Factual background & dispute
- A voice actor recorded audio for an audio production company; the authorization did not extend to AI training or voice synthesis.
- The audio recordings were processed using AI technologies into a voice product, which multiple companies subsequently sold, repackaged, and provided to end users.
- Users could input text to generate voice content replicating the voice actor's timbre, tone, and pronunciation style.
02
Core issues & judicial focus
- Whether AI-synthesized voices fall within the scope of legal protection for natural persons' voice rights
- Whether a standard audio recording license covers AI processing and commercial utilization
- Whether downstream purchasers of the voice product bear liability for damages
03
Judicial finding & holding
- The court held that identifiable AI-synthesized voices are protected under personal voice rights.
- The owner of the sound recordings had no authority to license AI processing of the claimant's voice, and the text-to-speech developer failed to conduct reasonable review of the authorization chain. They were held jointly liable for RMB 250,000 in economic damages.
- The AI dubbing platform and the text-to-speech developer were required to apologise. The dubbing platform and cloud-service distributors were not liable for damages because subjective fault was absent on the facts.
04
Practical risk implications
01Voice acquisition contracts must explicitly define the scope of AI training, synthesis, voice cloning, model deployment, and sublicensing.
02When procuring voice models, enterprises should verify the underlying chain of title, usage restrictions, and takedown mechanisms.
03Voice products should undergo identifiability testing before commercial launch and provide clear complaint channels for rights holders.