(Court) actions speak louder than words: China’s first court judgment on AI-generated voice rights infringement

Barely a month after China’s first copyright infringement case concerning AI-generated content, the Beijing Internet Court ruled on the country’s first case regarding the infringement of personality rights involving AI-generated voice. 

In its ruling, the Court found that a software company had infringed an individual's rights to likeness over their voice by using an AI-powered tool to mimic their voice without consent and distributing it to other platforms. 

Facts of the case

The claimant is a voice actor who discovered that their voice was being used without their consent across various applications. On further investigation, the claimant found that their voice had been reproduced by an AI text-to-speech tool using and modifying their previous voice recordings as the input source. These recordings were initially made by the claimant for a media company, which had granted the developer of the AI tool the rights to use the voice recordings.

Following this discovery, the claimant filed a suit against, among others, the media company and the AI developer, alleging that their actions had violated their voice rights. The defendants argued, among other things, that the AI-processed voice was not the same as the claimant’s original voice and could be distinguished. 

What rights does an individual have over their voice under Chinese law?

Chapter IV of the Chinese Civil Code provides an individual with the general right to likeness, allowing the individual to make, use, publicise or authorise others to use their image in accordance with the law. No organisation or third party may make, use or publicise the image of the rightsholder without their consent. 

These image rights are differentiated from the rights in the works encapsulating the image, such as a photograph or a video recording, and the Chinese Civil Code explicitly prohibits the rightsholder of the latter from using or publicising the image without the consent of the party holding the right to likeness. 

In determining that the claimant had a general right to likeness over their voice, the court referred to Article 1023 of the Chinese Civil Code, which provides that the protection of a natural person's voice should be treated the same as the protection of their likeness or image right. This afforded legal protection for the claimant’s voice and allowed the court to focus its analysis on whether the rights over the claimant’s voice could be extended to the AI-generated voice.

Does a person’s “voice right” extend to an AI-generated voice?

The court acknowledged that a natural person’s voice possesses distinct characteristics such as its “voiceprint, timbre and frequency”, all of which are unique to an individual. Given these distinctive features, the voice of a natural person can become identifiable after repetition or prolonged exposure. 

Taking the nature of an individual’s voice into consideration, the court ruled that, if an artificially synthesised voice can cause the general public to associate the voice with a particular natural person, then the voice is deemed identifiable.

Applying these principles, the court took the view that the AI-generated voices were highly consistent with the claimant’s voice timbre, intonation, and pronunciation style, which can cause the public to relate to, or attribute the AI-generated voices with the claimant’s. Accordingly, the court extended the claimant’s voice rights to the AI-generated voice in dispute.

Copyright and protecting personality rights

On the defence raised by the AI developer and media company holding the copyright to the claimant’s previous voice recordings, the court held that the relevant licensing arrangements do not include the rights to use the claimant’s voice for modification and redeployment by AI without the individual’s informed consent. 

As such, the court rejected the defence raised and found that the AI developer and the media company’s unauthorised use of the claimant’s voice constituted an infringement of their voice rights and allowed for compensation.

Global perspective

GenAI service providers that offer voice assistant features or text-to-speech functions in China and beyond should ensure that any voice adopted by AI applications do not infringe on any individual’s voice rights. 

The ruling comes at a time when the protection of the right to likeness and other personality rights is quickly gaining attention around the world, driven by the rapid development of voice replication devices and deep fakes more generally. For instance, in May, Scarlett Johansson criticised OpenAI for releasing an AI voice assistant that resembled her own voice, after she declined to grant a licence. 

In the US, the law in some states generally recognise rights of publicity and personality rights that protect against the unauthorised publication of individuals’ private information and commercial exploitation of aspects of their persona, including voice. 

By contrast, English law does not provide specific rights relating to publicity or personality, instead relying on other causes of action, such as the tort of passing off and claims under privacy laws. Other common law jurisdictions, such as Singapore and Hong Kong SAR, take a similar position. However, such causes of action are typically restricted to celebrities and famous individuals. 

A rapidly evolving legal landscape

These principles within existing intellectual property regimes and the more nascent GenAI regulatory landscape are rapidly evolving, both in China and globally. It will be imperative that GenAI service developers and providers keep abreast of these changes to allow them to seize opportunities that emerge as well as mitigate potential IP infringement risks.