Recently, the Cyberspace Administration of China has carried out the "Clear and Bright · Rectifying AI Application Chaos" special action in its second phase. The focus is on addressing prominent issues such as using AI to create and spread false information, spreading violent and vulgar content, impersonating others, and infringing on the rights of minors. Local cyberspace administrations have been deployed to advance key rectification tasks.
As of now, the special action has removed 5.61 million illegal and irregular pieces of information, investigated 49,000 accounts, and dealt with more than 2,400 websites and applications.
Platforms and Manufacturers Are Tightening Content Governance
During this period, website platforms have continuously improved their AI content governance mechanisms. Platforms such as Douyin, Kuaishou, Weibo, Tencent, Baidu, Bilibili, Xiaohongshu, Zhihu, Douban, and Taobao have optimized multimodal recognition models, dynamically expanded face libraries, voiceprint libraries, and violation sample libraries, and launched real-time face governance systems, among other measures. A total of 46 governance announcements have been issued. Platforms such as Tongyi, Yuanbao, Qianwen, and Wenyiyan have strictly reviewed original data and limited the output of illegal content, while also strengthening the implementation of requirements for identifying AI-generated synthetic content. App stores such as Huawei, Xiaomi, OPPO, and Vivo have established stricter developer access standards and listing requirements, and have conducted spot checks and retests on apps currently on the shelves.
Seven Typical Cases Have Been Exposed
During the special action, seven typical cases were exposed. First, AI was abused to "modify" classic works such as "Romance of the Three Kingdoms" and "Journey to the West," subverting important character images and fabricating low-quality, sensational plots to gain traffic, known as "digital garbage" content, involving multiple accounts on Weibo, Douyin, Kuaishou, and WeChat. Second, AI was used to generate false disaster scenes and rescue footage, fabricate false social hot topics, post feudal superstitious content, or use AI digital people to falsely promote and sell products, misleading the elderly. Third, AI was used to "swap faces" and "change voices" to impersonate public figures, generating vulgar dance images and videos, and kissing and hugging scenes for profit. Fourth, large-scale release of AI-generated content containing terrifying, weird, violent, and vulgar images. Fifth, posting AI-generated content about underage pregnancy and fighting that does not match age characteristics, mass-producing and spreading "toxic animations" and children's "torture" and "heretical" videos. Sixth,引流 for AI "management" software, simulating real person operations of accounts, commenting and interacting, and generating a large amount of low-quality, homogenized copy for online water army activities. Seventh, some websites and platforms' intelligent agent covers, descriptions, and generated dialogues contain obscene, pornographic, and harmful information to the physical and mental health of minors.
All the relevant accounts have been dealt with according to law and regulations; for websites and platforms that refuse to correct, measures such as warnings, fines, removal, or closure will be taken according to procedures. The cyberspace administration stated that in the next step, it will continue to closely monitor new situations and risks of AI-generated synthetic content and take timely effective measures to prevent the abuse of AI technology.