New rules set out accountability for deepfakes, voice cloning and the disclosure of personal information using artificial intelligence.
The Supreme People's Court of China released a guideline on Sept 7 to curb the misuse and abuse of artificial intelligence (AI) and bolster the protection of individuals' rights in the digital age, China Daily reports.
This marks the first set of judicial rules on the handling of AI-related cases to be released by the country's top court.
The 24-article document responds to hot-button issues that have stirred public debate over the past few years, including deepfakes, voice cloning and doxing — the practice of obtaining and publicly disclosing others' personal information online. The guideline provides criteria for identifying those who use AI to infringe on personality rights and holds them accountable.
According to the guideline, using someone's voice without consent to create synthetic speech that mimics their timbre, tone and speaking style constitutes an infringement of their voice rights.
Similarly, using AI to process a person's name or likeness without permission to create and distribute a recognizable digital avatar is considered a violation of their right to name and likeness.
Zhou Jiahai, head of the research office of the Supreme People's Court, noted that the basic aim of the new rules is to balance development with security.
"It strikes a balance between curbing AI abuses to strengthen judicial protection of people's rights in the digital era, and supporting technological and industrial innovation to safeguard the healthy development of AI,"
he said.
The office stressed that these rights are fundamental to personal dignity and rank among the most essential civil rights. Protecting them requires a tolerant yet prudent approach, taking into account the interests of the AI industry's development. At the same time, tolerance should not turn into permissiveness, nor should prudence mean acquiescence.
Modern AI technologies allow the likeness of virtually anyone to be used, making ordinary people potential targets of deepfakes, the court's research office said. According to experts, voices can also be "copied" using AI at very low cost, with results that can be virtually indistinguishable from the original.
These issues were among the key reasons behind the formulation of the new guideline. The rules are also intended to standardize judicial practices across the country and curb the use of AI to infringe on personality rights.
Chinese courts have already heard cases related to AI. In 2024, the Beijing Internet Court found AI voice imitation without the consent of its owner to be an infringement of personality rights. The case involved a voice-over artist whose recorded audiobooks were used without her consent to build a commercial AI text-to-speech service, which was then sold on other platforms. The court ordered the violator to pay compensation.
The guideline also calls for tougher measures against doxing, one of the most common forms of cyberbullying. Doxing seriously violates the right to privacy, disturbs people's daily lives and undermines security online, the research office of the Supreme People's Court said.
In response, the guideline clarifies that using AI to track and analyze publicly available information about individuals — such as phone numbers, online accounts or social media profiles — to obtain private information, "leak" it or publicly disclose it should be deemed an infringement of the right to privacy.
Furthermore, the guideline sets forth provisions concerning consumer rights protection, intellectual property protection and autonomous driving in the context of AI.
Given the rapid development of technology, some provisions of the document have deliberately been formulated in broad terms, experts noted. More specific rules will follow as judicial experience accumulates.
"Amid rapid technological change, we hope to draw clear legal boundaries in order to support innovation while ensuring that AI develops swiftly, soundly and steadily,"
the court said.
Source: China Daily
