Cyberspace Administration of China: Strictly Limit the Generation of Violating Content in platforms such as Douyin, Kuaishou, Zhihu, and Weibo

Dynamic Monitoring Beating AI Newsletter: The Cyberspace Administration of China recently launched the second phase of the special action "Clean and Rectify Disorderly AI Applications," focusing on rectifying issues such as the use of AI to generate and disseminate fake information, violent and vulgar content, impersonation of others, and infringement of the rights of minors. As of now, the special action has cumulatively cleaned up more than 5.61 million pieces of illegal and irregular information, investigated over 49,000 accounts, and dealt with more than 2,400 infringing websites and applications.

On the front of AI content governance, platforms such as Douyin, Kuaishou, Weibo, Tencent, Baidu, Bilibili, Xiaohongshu, Zhihu, Douban, and Taobao have further enhanced their multimodal recognition capabilities, dynamically expanded their facial recognition database, voiceprint database, and violation sample database, and intensified efforts to address illegal accounts and information, issuing a total of 46 governance notices.

At the same time, AI platforms such as Bean, Yuanbao, Qianwen, and Wenxin Yiyuan are strictly scrutinizing the review of original language data, strictly limiting the output of non-compliant content, and strengthening the requirements for identifying AI-generated synthetic content; Huawei, Xiaomi, OPPO, Vivo, and other app stores are enhancing the access and listing review for app developers, conducting spot checks and retesting of apps on the shelves.

During the special action, the Cyberspace Administration also reported multiple typical cases, including using AI to "modify" classic literary works, generate fake disaster and social hot topic information, AI face-swapping and voice-cloning impersonation of public figures, mass production of violent and vulgar content, dissemination of information that infringes the rights of minors through AI generation, and the use of AI "hosting" software for online water army activities. For non-compliant AI products and applications, relevant authorities will take actions such as warnings, fines, delisting, or shutdowns in accordance with the law.