The Cyberspace Administration of China has launched the first phase of the special campaign "Clean up and Rectify Irregularities in the Use of AI".

The special action of "Clear and Rectify AI Application Chaos" has been launched since April 2026. The Cyberspace Administration of China focuses on AI application chaos such as failure to fulfill the obligation of registering large-scale models, insufficient security and filtering capabilities of AI platforms, AI data poisoning, and inadequate implementation of synthetic content labeling. Local cyberspace administrations have deployed in-depth initiatives to promote key rectification tasks in the first phase, and have set up a "Report Area for AI Application Chaos" to accept public reports. With the joint efforts of all parties, in the first phase, over 14,000 violations of websites, applications, and AI products have been disposed of, over 6 million pieces of illegal and irregular information have been cleared, over 26,000 accounts have been handled, and over 1,300 illegal AI products and 9 illegal open-source data sets have been removed, showing positive progress in various aspects of work. Local cyberspace administrations have conscientiously performed their responsibilities for local management and taken effective supervision measures according to their actual conditions. For example, the Beijing Cyberspace Administration has established a mechanism for "platform self-inspection + regular inspection + technical screening + clue verification" to timely remove a batch of illegal AI applications and operating accounts. The Shanghai Cyberspace Administration has refined work requirements for platform types and guided platforms to upgrade their reporting mechanisms and proactively release governance notices. The Zhejiang Cyberspace Administration has organized targeted rectification activities from multiple dimensions such as model review capabilities, secure training data, and AI data poisoning, optimizing semantic judgment capabilities, and conducting high-frequency regression testing. The Jiangsu Cyberspace Administration has set up a special reporting window for AI application chaos, collected reports on 5 types of illegal and irregular situations such as AI data poisoning and failure to implement requirements for labeling generated synthetic content, and promptly verified and disposed of them. The Guangdong Cyberspace Administration has established a multi-department coordination and supervision mechanism for local areas, comprehensively using technical inspections, manual reviews, on-site supervision, and talk and punishment methods to carry out the full-chain governance of AI application chaos. Key website platforms actively fulfill their main management responsibilities and strengthen control measures in a targeted manner. For example, Huawei has added special audits and labeling for generative artificial intelligence services in its app store; Alibaba has improved its digital fingerprint comparison and keyword interception mechanisms to enhance the accuracy of identifying illegal and irregular content; KNOWLEDGE ATLAS has built a multimodal review model and conducted cross-review for complex scenarios such as intent reversal in multi-round dialogues; Xiyu timely identifies and blocks malicious activities such as creating viruses and phishing software using AI; DeepSeek embeds anomaly detection algorithms in data collection and preprocessing stages to prevent malicious sample manipulation, and so on. Next, the Cyberspace Administration of China will focus on prominent issues such as using AI technology to produce and disseminate false information, spread violent and vulgar content, impersonate others, infringe on the rights of minors, and engage in online water army activities, and will launch the second phase of the "Clear and Rectify AI Application Chaos" special action to increase the crackdown on illegal accounts and organizations, strengthen the management of key aspects of AI applications, urge platforms to enhance their AI prevention and control capabilities, and focus on maintaining a clean network ecology and promoting the positive development of artificial intelligence. This article was reprinted from the "Cyberspace China" WeChat public account, edited by GMTEight: Liu Jiayin.