In an announcement dated April 23, 2024, the Beijing Internet Court of China ruled against a defendant that used a dubber’s voice to train a generative AI. Specifically, the Court held that the defendant use of dubber’s voice to train an AI voice generator infringed the dubber’s personality rights due to similarity of the voice generator output to the dubber’s voice. This same Court earlier held there is copyright in AI-generated images.
As the Beijing Internet Court explained, the plaintiff Yin is a dubbing artist. After being informed by a friend, the plaintiff discovered that the works produced by others using his dubbing were widely circulated on many well-known apps. After sound screening and tracing, it was found that the sounds in the above-mentioned works came from a text-to-speech product on a platform operated by the defendant, a smart technology company in Beijing. Users can realize the function of converting text into speech by inputting text and adjusting parameters.
The plaintiff had been commissioned by the second defendant, a cultural media company in Beijing, to record sound recordings. The second defendant was the copyright owner of the sound recordings. Defendant No. 2 later provided the audio of the sound recordings recorded by the Plaintiff to Defendant No. 3, a software company, allowing Defendant No. 3 to use, copy, and modify the data for commercial or non-commercial purposes for its products and services. The third defendant only used the sound recordings recorded by the plaintiff as material for AI processing, generated the text-to-speech product involved in the case, and sold it on a cloud service platform operated by a Shanghai network technology company, the fourth defendant. The first defendant, a smart technology company in Beijing, signed an online service sales contract with the fifth defendant, a Beijing technology development company. The fifth defendant placed an order with the third defendant, including the text-to-speech products involved in the case. The defendant, a smart technology company in Beijing, used an application program interface to directly retrieve and generate text-to-speech products for use on its platform without technical processing.
The plaintiff claimed that the defendant’s behavior had seriously infringed upon the plaintiff’s personality right covering voice rights. The first defendant, a smart technology company in Beijing, and the third defendant, a software company, should immediately stop infringement and apologize, and the fifth defendant should compensate the plaintiff for economic losses and emotional distress.
The defendant 1, a smart technology company in Beijing, argued that as a bona fide third party, it did not participate in the authorization process and legally purchased the artificial intelligence synthesized sound product involved through formal channels. It had fulfilled its duty of reasonable care and therefore the acts did not constitute infringement. Further, the voice products synthesized by artificial intelligence are different from the voices of natural persons in terms of personality rights. Current technology will watermark the artificial intelligence synthesized voices, cut off the connection between the artificial intelligence synthesized voices and the voices of natural persons, and will not produce voices corresponding to natural persons.
Defendant 2, a cultural media company in Beijing, argued that its use of sound recordings with audio copyright and neighboring rights in cooperation with defendant 3, a software company, complied with relevant legal provisions, and that agreements had been signed and fees paid at all stages, and there was no ill intent of infringement. The dubbing content and scenes suitable for AI audio products are different from simple audiobooks.
The third defendant, a software company, argued that it had obtained authorization for the sound involved in the case and did not have any subjective fault for infringement. The voice involved in the case was an AI-enabled voice and was not identifiable with the plaintiff’s personality.
The fourth defendant, a Shanghai network technology company, argued that it was only a cloud service platform provider and had fulfilled its reasonable duty of care and should not bear infringement liability.
The fifth defendant, a technology development company in Beijing, argued that it was neither the producer nor the user of the AI sound involved in the case, and was unaware of the infringement and should not bear infringement liability.
The Court held that a natural person’s voice is distinguished by voiceprint, timbre, and frequency. It has the characteristics of uniqueness and stability. It can form or cause ordinary people to have thoughts or emotional feelings related to the natural person, and can display personal traits and identity the speaker to others. The recognizability of a natural person’s voice means that a specific natural person can be identified through the voice characteristics based on repeated or long-term listening by others. A voice synthesized using artificial intelligence can be deemed to be identifiable if the general public or the public in related fields can associate it with the natural person based on its timbre, intonation and pronunciation style.
In this case, the third defendant, a software company, used the plaintiff’s personal voice to develop the text-to-speech product involved, and after in-court inspection, it was found that the AI voice was highly consistent with the plaintiff’s timbre, intonation, pronunciation style, etc., and could arouse the attention of ordinary people. The AI-generated voice connects the voice to the plaintiff himself, thereby identifying the plaintiff. Therefore, the plaintiff’s voice rights extend to the AI voice involved in the case.
Defendant No. 2, a cultural media company in Beijing, and Defendant No. 3, a software company, cannot establish the defense that they obtained legal authorization from the plaintiff.
In this case, the second defendant, a cultural media company in Beijing, enjoys copyright and other rights to the sound recordings, but does not include the right to authorize others to use the plaintiff’s voice in AI. Defendant 2 signed a data agreement with Defendant 3, a software company, and authorized Defendant 3, a software company, to use the plaintiff’s voice in AI without the plaintiff’s informed consent, which has no legal source of rights. Therefore, the defense of the second and third defendants that they obtained the plaintiff’s legal authorization cannot be established.
The use of the plaintiff’s voice without permission constitutes infringement. Compensation for damages should comprehensively consider the circumstances of the defendant’s infringement, the value of similar market products and other factors.
The second defendant, a cultural media company in Beijing, and the third defendant, a software company, used the plaintiff’s voice in AI without the plaintiff’s permission, which constituted an infringement of the plaintiff’s voice rights. Their infringement caused damage to the plaintiff’s voice rights and interests, and they should bear the corresponding legal consequences. responsibility. Defendant 1, a smart technology company in Beijing, defendant 4, a network technology company in Shanghai, and defendant 5, a technology development company in Beijing, have no subjective fault and are not liable for damages. Therefore, damages will be determined based on comprehensive consideration of factors such as the defendant’s infringement, the value of similar market products, and the number of product broadcasts.
The first defendant, a smart technology company in Beijing, and the third defendant, a software company, should apologize to the plaintiff. The second defendant, a cultural media company in Beijing, and the third defendant, a software company, must compensate the plaintiff for losses totaling 250,000 RMB.
The original text of the announcement is available here (Chinese only).
