China: New Ruling on Non-Infringement of Personality Rights by AI Voice: Does It Signal a Shift in Protection Standards?

As generative AI content becomes increasingly prevalent, concerns grow over whether personal attributes like voices and facial features can be protected from infringement and misuse (e.g., for telecom fraud), with related disputes continuing to multiply.

In April 2024, the Beijing Internet Court issued China’s first ruling on AI voice infringement [Case No.: (2023) Jing 0491 Min Chu No. 12142]. This legally effective ruling affirmed that voices recognizable by relevant public are protected under personality rights. Commercial AI utilization requires specific, explicit authorization from rights holders—general copyright licenses for voice usage are insufficient. In this case, the defendant recorded the plaintiff’s original voice, processed it with AI without authorization, and generated commercial products. The court ultimately determined this constituted infringement of the plaintiff’s personality rights. For detailed analysis,  see our earlier article China’s First AI Voice Personality Rights Infringement Rulings and Implications.

Notably, in December 2024, the Chengdu Railway Transport No. 1 Primary Court reached a contrary conclusion in another AI voice infringement case—the "Meng Shuai" virtual voice package dispute [Case No.: (2024) Chuan 7101 Min Chu No. 8969]^*1. The judgment deemed the AI voice non-infringing of personality rights. This ruling has taken effect and was recently made public.

These diametrically opposed outcomes raise a critical question: Has the standard for determining voice rights infringement changed?

"Meng Shuai" Case Summary

The plaintiff is a professional voice actor possessing commercial value and stable partnerships in fields like advertising, audiobooks, and corporate videos—though not a publicly recognized "celebrity voice actor." The plaintiff claimed distinctive recognizability for his voice among relevant public (e.g., advertisers and production companies).

Upon discovering that the "Meng Shuai" AI voice package in the defendant’s "Super Voice App" generated sounds highly resembling his own voice, the plaintiff filed suit demanding cessation of personality rights infringement and compensation. The plaintiff submitted four voiceprint comparison reports indicating 88.32%-95.35% similarity between his original voice and the "Meng Shuai" AI output.

The defendant argued that its AI voice model synthesized new voices using open-source or legally authorized training data—not by cloning any individual’s voice. They emphasized the plaintiff’s lack of public recognition, contending that ordinary users would not associate the "Meng Shuai" voice with the plaintiff or cause confusion. The defendant also provided two voiceprint reports showing merely 65.36% and 63.46% similarity.

After in-court auditory comparison of voice samples, the Chengdu court acknowledged similarities but found neither voice sufficiently recognizable to cause the "general public" to associate the AI output with the plaintiff or mistake its origin. The court declined to admit the voiceprint evidence as probative, further stating that absent proof of infringement, the plaintiff should tolerate "similarity of voices within reasonable limits." Consequently, all claims were dismissed. With neither party appealing, the judgment became final.

Comparison with China’s First AI Voice Infringement Case

Regarding training data sources - In the first AI Voice case, the defendant admitted exclusively using the plaintiff’s voice samples for AI training. In the Meng Shuai case, the defendant claimed utilization of mixed open-source or lawfully authorized data. The court did not compel disclosure of specific training data sources, leaving factual clarification on whether the plaintiff’s voice was actually used unresolved.

Concerning the association between the accused AI voice and the plaintiff - The first AI Voice case faced no dispute regarding personal association due to the defendant’s admission. In the Meng Shuai case, the court rejected voiceprint evidence demonstrating over 90% similarity, denying a strong personal association between the AI voice and the plaintiff. Technically, human voices possess uniqueness, singularity, and stability through characteristics like voiceprints, timbre, and frequency. Voiceprint analysis conclusions are typically regarded as objective evidence for determining personal attribution.

Regarding the recognizability standard - The first AI Voice case adopted the criterion of "whether ordinary listeners in relevant fields can recognize" the voice (e.g., discernible by professional groups like advertisers or production companies). The Meng Shuai case employed a "must be recognizable by the general public" standard (requiring celebrity-level public awareness).

Additionally, the Meng Shuai case introduced a "duty of tolerance," asserting that voice similarity arising from technology is unavoidable and the plaintiff must tolerate "similarity of voices within reasonable limits." The first AI Voice case did not mention this duty, focusing instead on establishing infringement criteria.

Controversies and Implications of the "Meng Shuai" Case

The judgement, when made public, sparked a number of questions^*2*3.

Firstly, the "Basic Security Requirements for Generative Artificial Intelligence Services" (TC260-003)^*4 issued by the National Technical Committee on Cybersecurity of Standardization Administration of China on 29 February 2024 require enterprises to identify infringement risks in training corpora prior to AI training. In this case, however, the court did not compel the defendant to disclose data sources, effectively placing the absolute burden of proving "data infringement" solely on the plaintiff.

Secondly, the technical conclusion of over 90% similarity of voiceprints as biometric evidence with uniqueness and stability was not adopted, leading to a lack of objective evidence to support the court's finding in this case that the AI voice at issue did not have identifiable personal attributes.

Thirdly, the court raised the standard of voice identifiability to a level of fame that the general public needs to be aware of, ignoring the voice-over industry's reliance on specialised identification by vertical sectors (e.g. advertisers, production companies).

Furthermore, the court failed to define the boundaries of "reasonable limits" while imposing a "duty of tolerance" on the plaintiff regarding voice similarity.

Some of the more direct criticisms include, for example, the court's failure to balance technological development and rights protection in applying the principle of “technology neutrality”, allowing enterprises to plead “technological confidentiality”, which in effect removes the burden of proof from enterprises, and eroding the boundaries of the protection of individuals' rights by imposing a duty of tolerance in favour of enterprises, and raising the prerequisites for voice protection to the standard of “recognisability to the general public”, which would essentially deprive non-celebrity voice actors of the right to protect the commercial value of their voices, and would expose most voice actors to the risk of having their voice assets harvested by technology for free.

It is worth noting that in June 2025, China's first AI Voice Infringement case was selected as the Supreme People's Court's “Typical Case of Infringement of Personality Rights through the Use of Networks and Information Technology”^*5. Does this signify the value orientation of the highest judicial authority towards similar cases?

References:

1. Intellectual Property Library (知产库), "Voiceover Artist v. AI Voice Big Model Infringement Judgment (配音员诉AI语音大模型侵权案判决书)" (in Chinese), https://mp.weixin.qq.com/s/MG3hhbRpfh3_06S5_t3gYQ.

2. Shang Jianguang (商建刚), "Voiceover Artist v. AI Voice Big Model Infringement Case Commentary (配音员诉AI语音大模型侵权案评析)" (in Chinese), https://mp.weixin.qq.com/s/yIwVt-cc4jul1CyqNm1qYw;

3. Qian Tao, Lawyer, and His Team (钱涛律师团队), "AI Voice-Over in Lawsuit:" Whose Voice Is ‘Meng Shuai’ Imitating? Legal Protection of Voice Rights and Interests from an AI Voice Infringement Case (AI配音惹上官司:“孟帅”是不是在模仿谁的声音？从一起AI语音侵权案看声音权益的法律保护)" (in Chinese), https://mp.weixin.qq.com/s/YCQvcPMSuM7RsmI0iG2ISA.

4. https://www.tc260.org.cn/upload/2024-03-01/1709282398070082466.pdf

5. https://www.court.gov.cn/zixun/xiangqing/467511.html