Beijing Internet Court finds an AI text-to-speech voice made from a voice actress's recordings was sold without her consent and awards her 250,000 yuan

The voice. According to the Beijing courts' case digest (published in the "Jing Dian" case column and entered in the People's Court case library), the voice actress recorded a sound recording for a Beijing media company in 2019. The media company, which held the copyright in the recording, signed a data-licensing agreement with a software company on 10 May 2019 and supplied a data authorisation that the voice actress had not signed. The software company used only her one recording as material, processed it with AI and other techniques into a text-to-speech product, and sold it on a cloud-service platform run by a Shanghai company. A Beijing app operator bought the product through another company and called it through an interface for use on its platform, which showed the voice as played 3,256,728,530 times.

The discovery and claim. In May 2023 she found that works using her voice were circulating widely on several well-known apps, and traced the voice to the text-to-speech product. She sued five companies, asking for 500,000 yuan for economic loss and 100,000 yuan for mental harm. The app operator and the software company took the product down during the proceedings.

The ruling. At the hearing the court used the product to convert text and found the AI voice highly consistent with her timbre, intonation and pronunciation style. On 23 April 2024 it held that a natural person's voice rights cover an AI-synthesised voice that the public can identify with that person, that the media company's rights in the recording did not include authorising AI use of her voice, and that the media company and the software company had infringed. It ordered written apologies and 250,000 yuan in economic damages, found the other three companies not at fault, and rejected the mental-harm claim because she had not shown serious mental harm. Neither side appealed.

Limits. All four sources relay the court's own account; the judgment text was not inspected. The parties are anonymised. The reports do not say when the product was first sold or used.

According to the Beijing courts' account, an AI text-to-speech voice made from a voice actress's recording without her consent was sold and widely used, and the court found her voice rights infringed and awarded 250,000 yuan for economic loss.

Beijing Internet Court civil judgment (2023) Jing 0491 Min Chu No. 12142 of 23 April 2024: written apologies from the app operator and the software company; 250,000 yuan for economic loss payable jointly by the media company and the software company; other claims, including 100,000 yuan for mental harm, rejected. The judgment took effect without appeal. The two companies took the product down during the first-instance proceedings (Beijing courts' case digest).