China: Court Protects Voice Right and Curbs AI Misuse
Published 4 June 2025
Fei Dang
On May 27, 2025, the People’s Supreme Court (the SPC) issued a series of typical cases regarding the implementation of the Civil Code. Among them, there is a case of infringing on the personality rights that is worthy of attention, as it draws a line between protecting voice right and restricting AI misuse.
The Plaintiff is a voice artist who found others had used her voice works widely broadcasted in multiple famous APPs. It is found that the voice in the said works came from the text-to-speech product in a platform operated by an intelligence technology company (the Defendant 1). According to the court examination, the Plaintiff used to make a sound recording upon entrustment from a cultural company (the Defendant 2) that is the copyright owner of the said sound recording. Then the Defendant 2 signed a data authorization use agreement with a software company (the Defendant 3), which provided the recording data, including the ones made by the Plaintiff to the Defendant 3. The Defendant 2 also provided a data authorization that was not signed by the Plaintiff to the Defendant 3. Then the Defendant used one of the recordings made by the Plaintiff as material, proceeded with AI technology, generated the text-to-speech products involved in the case, and sold it on a cloud service platform operated by a network technology company (the Defendant 4). The Defendant 1 signed an online purchase agreement with a technology company (the Defendant 5), which provided that the Defendant 5 shall make a purchase from the Defendant 3 including the said text-to-speech product. Then the Defendnat 1 took the form of an application program interface to directly access and generate the text-to-speech product for use in their platforms without technical processing. It is shown in the platform that the involved voice had been played 3,256,728,530 times.
The Plaintiff considered such action had severely violated her voice right and requested the court to order the Defendants 1 and 3 immediately cease the infringement and apologize and all five Defendants to compensate her for her economic loss RMB 500,000 yuan plus mental damages RMB 100,000 yuan. During the first instance trial, the Defendants 1 and 3 removed the involved text-to-speech product.
Upon using the involved product to conduct a text-to-speech operation in an on-the-spot investigation during the trial, it is found that the obtained AI Voice is highly in consistence with the Plaintiff’s tone color, intonation, pronunciation style, and so on. Accordingly, the Beijing Internet Court delivered the (2023) Jing 0491 Min Chu No. 12142 Civil Judgment and ordered that 1. The Defendants 1 and 3 apologize to the Plaintiff in writting within seven days of the effective date of the judgment; 2. The Defendant 2 and 3 jointly compensate the Plaintiff economic loss RMB 250,000 yuan; 3. Reject the Plaintiff’s other requests. None of the parties appealed, and the said judgment has come into effect.
According to the court, there were three key issues in the case: 1. Whether the Plaintiff’s voice right includes the voice processed by AI; 2. Whether the use of the Plaintiff’s voice in the case had obtained legal authorization; 3. Whether the claimed action had constituted infringement and what kind of liabilities to be borne if it did.
Regarding the key issue 1, the court recognized that the Plaintiff’s voice right shall include the voice generated by AI. In accordance with Article 1023.2 of the Civil Code, “For the protection of a natural person’s voice, the relevant provisions on the protection of the right to likeness shall be applied mutatis mutandis.” Thus, the prerequisite for the legal protection of voice right is that it should be identifiable, which means that upon others’ listening repetitively multiple times or for a long time, it can identify certain natural persons by such voice characteristics. Whether a voice processed by AI can fall into the protection scope of the natural person’s voice right lies in the fact of whether it can identify certain natural persons via the said voice. If an AI generated voice can be associated with certain natural persons by the general public or the public in the relevant field based on its tone color, intonation, and pronunciation style, then it can be recognized that it is identifiable. As it is mentioned above, the AI Voice herein was highly consistent with the Plaintiff’s voice upon the on-the-spot investigation, and therefore the AI Voice belonged to the Plaintiff’s voice right.
Regarding the key issue 2, although the Defendant 2 is entitled to the copyright of the said sound recording, it excludes the right to authorize other to use the Plaintiff’s voice to be processed by AI and used. Since the said data authorization was not signed by the Plaintiff, the agreement signed between the Defendants 2 and 3 lacked the legal ground to use AI to process the Plaintiff’s voice without her consent.
As to the key issue 3, all five Defendants infringed the Plaintiff’s voice right and should bear liabilities respectively. Among them, the Defendants 2 and 3 processed the Plaintiff’s voice with AI and used the processed voice without authorization, which had constituted infringement to the Plaintiff’s voice and caused damages and therefore shall be liable to compensation, as well as an apology in writting. The Defendants 1, 4, and 5 shall not be liable for compensation as they lacked subjective fault. As to the mental damages, since the Plaintiff failed to prove that the infringement had caused severe damages to her mental interest, such request was not supported.
This case is quite enlightening on the grounds that 1) the PRC Civil Code incorporates the voice right into its protection, which applies to mutatis mutandis protection of the right to likeness (Article 1023.2), whereas the current case shows exactly how the said Article applies. 2) This case incorporates the AI-generated voice into the scope of protection of the voice right under the Civil Code on the premise that such AI voice can be identified as the voice of a certain natural person by the general social public or the public within a certain scope based on its tone color, intonation, and pronunciation style. Such interpretation and application of the law do not only reflect the impact of the latest technology development on personal right protection but also show that the legal protection on personal rights is dynamic. 3) It draws a legal line between the personal rights and interests protection and the reasonable use of AI technology in order to prevent the abuse of personal rights due to the rapid development of technology. There is no doubt that such a case will provide clear guidance on better regulating AI development without sacrificing personal rights.
The Plaintiff is a voice artist who found others had used her voice works widely broadcasted in multiple famous APPs. It is found that the voice in the said works came from the text-to-speech product in a platform operated by an intelligence technology company (the Defendant 1). According to the court examination, the Plaintiff used to make a sound recording upon entrustment from a cultural company (the Defendant 2) that is the copyright owner of the said sound recording. Then the Defendant 2 signed a data authorization use agreement with a software company (the Defendant 3), which provided the recording data, including the ones made by the Plaintiff to the Defendant 3. The Defendant 2 also provided a data authorization that was not signed by the Plaintiff to the Defendant 3. Then the Defendant used one of the recordings made by the Plaintiff as material, proceeded with AI technology, generated the text-to-speech products involved in the case, and sold it on a cloud service platform operated by a network technology company (the Defendant 4). The Defendant 1 signed an online purchase agreement with a technology company (the Defendant 5), which provided that the Defendant 5 shall make a purchase from the Defendant 3 including the said text-to-speech product. Then the Defendnat 1 took the form of an application program interface to directly access and generate the text-to-speech product for use in their platforms without technical processing. It is shown in the platform that the involved voice had been played 3,256,728,530 times.
The Plaintiff considered such action had severely violated her voice right and requested the court to order the Defendants 1 and 3 immediately cease the infringement and apologize and all five Defendants to compensate her for her economic loss RMB 500,000 yuan plus mental damages RMB 100,000 yuan. During the first instance trial, the Defendants 1 and 3 removed the involved text-to-speech product.
Upon using the involved product to conduct a text-to-speech operation in an on-the-spot investigation during the trial, it is found that the obtained AI Voice is highly in consistence with the Plaintiff’s tone color, intonation, pronunciation style, and so on. Accordingly, the Beijing Internet Court delivered the (2023) Jing 0491 Min Chu No. 12142 Civil Judgment and ordered that 1. The Defendants 1 and 3 apologize to the Plaintiff in writting within seven days of the effective date of the judgment; 2. The Defendant 2 and 3 jointly compensate the Plaintiff economic loss RMB 250,000 yuan; 3. Reject the Plaintiff’s other requests. None of the parties appealed, and the said judgment has come into effect.
According to the court, there were three key issues in the case: 1. Whether the Plaintiff’s voice right includes the voice processed by AI; 2. Whether the use of the Plaintiff’s voice in the case had obtained legal authorization; 3. Whether the claimed action had constituted infringement and what kind of liabilities to be borne if it did.
Regarding the key issue 1, the court recognized that the Plaintiff’s voice right shall include the voice generated by AI. In accordance with Article 1023.2 of the Civil Code, “For the protection of a natural person’s voice, the relevant provisions on the protection of the right to likeness shall be applied mutatis mutandis.” Thus, the prerequisite for the legal protection of voice right is that it should be identifiable, which means that upon others’ listening repetitively multiple times or for a long time, it can identify certain natural persons by such voice characteristics. Whether a voice processed by AI can fall into the protection scope of the natural person’s voice right lies in the fact of whether it can identify certain natural persons via the said voice. If an AI generated voice can be associated with certain natural persons by the general public or the public in the relevant field based on its tone color, intonation, and pronunciation style, then it can be recognized that it is identifiable. As it is mentioned above, the AI Voice herein was highly consistent with the Plaintiff’s voice upon the on-the-spot investigation, and therefore the AI Voice belonged to the Plaintiff’s voice right.
Regarding the key issue 2, although the Defendant 2 is entitled to the copyright of the said sound recording, it excludes the right to authorize other to use the Plaintiff’s voice to be processed by AI and used. Since the said data authorization was not signed by the Plaintiff, the agreement signed between the Defendants 2 and 3 lacked the legal ground to use AI to process the Plaintiff’s voice without her consent.
As to the key issue 3, all five Defendants infringed the Plaintiff’s voice right and should bear liabilities respectively. Among them, the Defendants 2 and 3 processed the Plaintiff’s voice with AI and used the processed voice without authorization, which had constituted infringement to the Plaintiff’s voice and caused damages and therefore shall be liable to compensation, as well as an apology in writting. The Defendants 1, 4, and 5 shall not be liable for compensation as they lacked subjective fault. As to the mental damages, since the Plaintiff failed to prove that the infringement had caused severe damages to her mental interest, such request was not supported.
This case is quite enlightening on the grounds that 1) the PRC Civil Code incorporates the voice right into its protection, which applies to mutatis mutandis protection of the right to likeness (Article 1023.2), whereas the current case shows exactly how the said Article applies. 2) This case incorporates the AI-generated voice into the scope of protection of the voice right under the Civil Code on the premise that such AI voice can be identified as the voice of a certain natural person by the general social public or the public within a certain scope based on its tone color, intonation, and pronunciation style. Such interpretation and application of the law do not only reflect the impact of the latest technology development on personal right protection but also show that the legal protection on personal rights is dynamic. 3) It draws a legal line between the personal rights and interests protection and the reasonable use of AI technology in order to prevent the abuse of personal rights due to the rapid development of technology. There is no doubt that such a case will provide clear guidance on better regulating AI development without sacrificing personal rights.