In April 2024, the Beijing Internet Court issued a first-instance judgment in China’s first “AI voice infringement case,” clarifying that the rights and interests of natural persons' voices extend to AI-generated voices and defining the boundaries of rights and infringement standards. The judgment, has taken effect, serving as a critical reference for similar cases. As disputes over AI-generated content increase, it is expected that the relevant legislative process in China will be pushed forward. At the current policy level, China’s National Radio and Television Administration has required generative AI content to be prominently labelled and reviewed as of December 2024, but legislative protections still need to be deepened.
Case Summary
Plaintiff is a professional voiceover artist. Defendant 2 (Company A) commissioned Plaintiff to record her original voice and then licensed it to Defendant 3 (Company B) for AI processing and generation of a commercial product, which was sold through Defendant 4's (Company C) cloud platform. Defendant 5 (Company D) purchased the product for Defendant 1 (Company E), which was actually used in the operation of Defendant 1. Plaintiff claimed that the unauthorised use of her voice by the five defendants constituted infringement to her personality rights, and demanded cessation of the infringement and compensation for damages. Case reference (2023) Jing 0491 Min Chu No. 12142.
Key Issues
- Are AI-generated voices protected by personality rights?
- Can the copyright licence for sound recordings cover AI-based voice usage?
- If an infringement exists, how is the liability of each defendant for the infringement divided?
Court Rulings
1. Extended protection of personality rights in the voice
The court held that the natural person's voice interest falls within the scope of personality rights. If an AI-generated voice’s timbre, intonation, style or other characteristics are highly consistent with those of a natural person, sufficient to enable the relevant public to identify a particular person, the voice is protected by personality rights. In this case, the AI voice was deemed identifiable with Plaintiff and was therefore protected by law.
2. Separation of copyright and personality Rights
The court held that although Defendant 2 owned the copyright in Plaintiff’s sound recordings, this right did not include the authorisation of the AI use of Plaintiff’s voice, and that Defendant 3’s AI processing based on Defendant 2’s permission were also without Plaintiff’s separate express consent. The court emphasised that the personality rights in the voice are independent of copyright, and that the voice in sound recordings cannot be commercially exploited by AI without express consent.
3. Limited liability of the parties
Based on the subjective fault and the linkage of the acts, the court ruled that Defendant 2 (the copyright owner of the sound recording) and Defendant 3 (the technical processor) were jointly liable for the infringement, and ordered an apology and compensation of RMB 250,000 (circa. US$35,000); Defendant 1 (the actual user of the AI voice) was required to apologise due to invalid procurement defences, but was not required to pay damages; and Defendant 4 (the cloud service provider) and Defendant 5 (the procurement intermediary) bore no liability due to lack of fault.
None of the parties appealed the above judgement.
Highlights
In this, the judgement addressed the challenges posed by AI technology to the traditional legal system, and for the first time extended the protection of personality rights to the field of virtual sound synthesis, and established the principle that the application of technology should not circumvent the authorisation of personality rights and the following criterions:
- Identifiability standard: adopt ‘whether ordinary listeners in the relevant field can identify (the voice)’ as the core test for determination, providing an objective yardstick for whether the AI sound in the technical scenario infringes on the personality right, and avoiding ambiguities in the judgement caused by the complexity of the technology.
- Authorisation separation criteria: clearly specify the principle of separation of authorisation of copyright and personality right, and require AI commercial use of voice to obtain the explicit consent of the right holder, so as to avoid the legal risk caused by blanket authorisation.
- Principle of responsibility hierarchy: distinguish direct infringement from indirect liability based on roles in the technology chain, balancing the needs of technological innovation and rights protection.
Implications
The major takeaway from this case for businesses is that when using AI sound technology, it is important to build a dual consent path of ‘original authorisation + specific authorisation for AI use’ to avoid legal risks related to personality rights.
