The Cyberspace Administration of China is carrying out the first phase of the special campaign "Clean and Rectify the Abuse of AI Technology" intensively.

The special action of "Clear and Rectify the Abuse of AI Technology" has been launched since April 2025. The Cyberspace Administration of China has focused on the chaotic situation of AI technology abuse, such as AI face-swapping and voice mimicry infringing on public rights, and the lack of AI content labeling misleading the public. It has carried out the key rectification tasks of the first phase, deployed local cyberspace departments to strengthen the disposal of illegal AI products, cut off the marketing channels of illegal products, urge key website platforms to improve technical security measures, and accelerate the implementation of content labeling for generated synthetic content. In the first phase, more than 3,500 illegal AI products such as mini-programs, applications, and chatbots have been disposed of, over 960,000 pieces of illegal information have been cleaned up, and more than 3,700 accounts have been disposed of, with positive progress in all work. Local cyberspace departments have taken multiple measures to solidly promote and actively fulfill their responsibilities for territorial management. The Beijing Cyberspace Administration has established a reporting channel for the governance of AI technology abuse, forming a working mode of "user tagging-platform verification-joint disposal" and handling 26 related reports. The Shanghai Cyberspace Administration has promoted the implementation of labeling requirements, organized four sessions of regulations dissemination and thematic exchanges, covering more than 400 companies, and key companies in the area have basically completed the specification of explicit labeling. The Zhejiang Cyberspace Administration has increased supervision of local AI applications and website platforms, with all platforms intercepting and cleaning up more than 25.5 million pieces of various illegal information and queries. The Jiangsu Cyberspace Administration has strengthened overall coordination, cooperated with the provincial communications management department to investigate 163 AI risk domain names, and blocked and canceled access to 18 domain names. The Tianjin Cyberspace Administration has used a "human+technology" approach to conduct large-scale security monitoring and disclosed four batches of 14 risk issues. Key website platforms actively fulfill their main management responsibilities and strengthen the governance of the source of AI technology abuse. In terms of disposing of illegal AI products, Tencent has standardized application management, raised entry barriers, optimized inspection mechanisms, rejected and disposed of a total of over 570 illegal mini-programs and applications. In terms of cleaning up illegal AI product tutorials and merchandise, Weibo has reviewed and disposed of over 4,800 pieces of illegal content through strategy identification, user reports, and other channels, and published typical cases. In terms of strengthening training corpus management, Tongyi Platform has established a secure management system around the data lifecycle, enhancing training corpus management in data collection, training, and use stages. In terms of enhancing security management measures, Douyin has established a "red-blue confrontation" mechanism, simulated attack cases, repaired potential security vulnerabilities, and optimized the model's ability to identify false information. In terms of implementing content labeling requirements, key platforms such as Alibaba, Kuaishou, and Xiyu actively promote the implementation of implicit metadata labeling. In terms of preventing security risks in key areas, Xiaohongshu inputs professional domain data in the post-training stage of the model to enhance the model's understanding of key issues in fields such as healthcare, finance, and minors. Next, the Cyberspace Administration of China will focus on addressing outstanding issues such as AI rumors and vulgar content in the second phase of the special action of "Clear and Rectify the Abuse of AI Technology." It will build a technical monitoring system, establish norms for disposal and punishment, promote the timely implementation of content labeling, form a long-term effective working mechanism, and strive to maintain a clean network ecosystem and promote artificial intelligence towards goodness and positivity. This article is reproduced from the "Cyberspace China" public account, GMTEight editor: Jiang Yuanhua.