From parodying entrepreneur Lei Jun to creating audio and videos of Dr. Zhang Wenhong to promote their own products, and even criminals using deepfake videos for fraud, the past year has seen a worsening trend of infringement through deepfake audio and video content.

Previously, some netizens had used fabricated recordings of Andy Lau's voice to gain online attention, prompting Andy Lau's film company to urgently issue a statement warning the public not to fall for scams involving voice synthesis technology that mimics his voice.

A recent investigation by journalists has revealed that creating deepfake audio and videos is not particularly difficult. On some e-commerce platforms, deep synthesis technology has even become a tool for many online stores to profit. For just a few dozen yuan, one can commission customized deepfake content featuring celebrities.

Online streamers teach how to use open-source tools for deep synthesis in live broadcasts. How are these deepfake voices and face swaps created? Journalists found that the corresponding deep synthesis software is not difficult to download online, and some streamers even host live sessions to teach netizens how to use it.

A reporter searched for face-swapping and voice-changing tools in a mobile app store and found numerous options. However, the reporter noticed that these face-swapping tools were limited to fixed scenarios and lacked high realism in terms of effect. So, how do online shops achieve highly realistic deepfake voices and face swaps?

Xue Zhihui, an expert from the Artificial Intelligence Security Governance Professional Committee of the China Cyberspace Security Association: Currently, there are many open-source software and tools available that can achieve effects like voice and face swapping, which netizens can freely download, install, and use on their own computers.

Experts say there are currently dozens of such software programs available online. However, whether it's voice-changing or face-swapping software, different scenarios require adjusting corresponding parameters, involving multiple steps. These are semi-professional tools with a certain technical threshold, which is why they are not widely known among netizens. Nonetheless, journalists have found that many content creators on short-video platforms are teaching netizens how to use such software.

Unauthorized synthesis of another person's video or voice constitutes infringement. So, does using deep synthesis technology on online shopping platforms to create and sell videos or audio of celebrities violate relevant laws? Experts state that related laws and regulations clearly stipulate that without authorization, one must not produce or publish deep-synthesized content of others. Online shops accepting orders to deep-synthesize celebrity videos are already engaging in infringing activities and should bear corresponding legal responsibilities.

Lv Yanhui, Member of the Computer Security Professional Committee of the China Computer Federation: The Civil Code of the People's Republic of China stipulates that natural persons enjoy rights and interests such as portrait and voice, and any organization or individual using information technology to forge them constitutes an infringement. The "Regulations on the Management of Deep Synthesis in Internet Information Services," jointly issued by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the Ministry of Public Security, require that no organization or individual may use deep synthesis services to produce, copy, publish, or disseminate information prohibited by laws and administrative regulations. The "Measures for the Identification of AI-Generated and Synthesized Content (Draft for Comments)," released by the Cyberspace Administration of China, also proposes that generated and synthesized content should be labeled to enable public identification and differentiation, preventing the spread of misleading information.

What legal responsibilities should online stores bear for using deep synthesis technology to create videos of celebrities? Zhao Jingwu, Associate Professor at the School of Law, Beihang University: In judicial practice, China has concluded its first case of personality rights infringement involving generated voices. In this case, the plaintiff, a voice actor, had his voice generated and sold without authorization. The court ruled that China's Civil Code, under the section on personality rights, recognizes a natural person's voice as a personal right with exclusive personal attributes. Therefore, the court ultimately determined that the defendant's actions constituted an infringement.

Expert: Raising Awareness and Utilizing Detection for Prevention In fact, it’s not just celebrities—ordinary people or even animated characters can be victims. If synthetic audio or video is created without the consent of the individual or copyright holder, it may be deemed an infringement, and the perpetrator could face legal consequences. As these technologies become more widespread, how can we guard against the risks of deepfake audio and video violations? Let’s hear the expert’s insights.

Xue Zhihui, Expert at the AI Security Governance Committee of the China Cyberspace Security Association: From a technical standpoint, we can actually use technology to counteract and detect. Currently, there are some typical technologies that can examine these images or audio-visual materials to determine or analyze whether they have been post-processed or synthesized.

Lv Yanhui, Member of the Computer Security Professional Committee of the China Computer Federation: Preventing infringement issues requires comprehensive measures at multiple levels, including legal, platform, and public aspects. At the legal level, it is necessary to further improve relevant legislation, refine legal provisions on cloning technology, clarify the definition and responsibilities of infringing acts, and strengthen law enforcement. At the platform level, efforts should be made to enhance the promotion and implementation of relevant policies and regulations, protect existing data, apply technical means to avoid cloning infringement issues, and establish sound content review and infringement reporting mechanisms to promptly identify and address infringement problems.

Legal experts also caution deep synthesis creators and information publishers against harboring any illusions of leniency, emphasizing that there are no gray areas in the law and warning them not to risk significant losses for minor gains.

Zhao Jingwu, Associate Professor at the Law School of Beihang University: The law does not prohibit the use of synthetic technology, but rather prohibits illegal and unreasonable use, especially the publication and dissemination of synthetic information without any clear labeling or disclosure.