According to the news of the public number of "net letter China", "clear and rectify". AI Since the launch of the "Technology Misuse" initiative in April 2025, the central government has been working on a number of initiatives.Internet Information OfficeFocusing on AI face-switching voice infringement of public rights and interests, the lack of AI content labeling to mislead the public and other misuse of AI technology, in-depth promotion of the first phase of the key remediation tasks, the deployment of the local Internet information departments to increase the disposal of non-compliant AI products, cut off the flow of illegal product marketing channels, supervise the key websites and platforms to improve the technical safety and security measures, and to promote the accelerated landing of the generation of synthetic content labeling.In the first phase, it disposed of more than 3,500 illegal small programs, apps, smart bodies and other AI products, cleaned up more than 960,000 pieces of illegal information, and disposed of more than 3,700 accounts.The work is progressing positively.
The local Internet information departments have taken many measures to make solid progress and actively fulfill their local management responsibilities:
- The Beijing Internet Information Office opened a reporting channel for the governance of AI technology abuse, formed a working model of "user marking - platform verification - joint disposal", and disposed of 26 relevant reports.
- The Shanghai Internet Information Office has promoted the marking requirements for landing demonstration, organized 4 sessions of regulations and thematic exchanges, covering more than 400 enterprises, and the key enterprises in the area have basically completed the specification of visible marking on line.
- Zhejiang Internet Information Office has increased its supervision of local AI applications and website platforms, and the platforms have intercepted and cleaned up more than 25.5 million pieces of all kinds of illegal and irregular information and questions.
- The Jiangsu Net Information Office has strengthened the coordination and united with the Provincial Communications Bureau and other departments to identify 163 AI risk domain names, and blocked and canceled access to 18 domain names.
- Tianjin Internet Information Office uses "artificial + technology" approach to carry out large model security monitoring, notified four batches of 14 risk issues.
Key websites and platforms have actively fulfilled their main management responsibilities and strengthened the governance of AI technology abuse at the source:
- In terms of disposing of non-compliant AI products, theTencent standardizes app management, raises access thresholdIn addition, it optimized the inspection mechanism and rejected and disposed of a total of more than 570 non-compliant applets and applications.
- In terms of cleaning up offending AI product tutorials and merchandise, theWeibo audits through multiple channels including policy recognition and user reportingIn addition, we have disposed of more than 4,800 pieces of non-compliant content and published typical cases.
- In the context of strengthening the management of the training corpus.Tongyi Platform establishes a security management system around the data lifecycle, strengthening training corpus management at the stages of data collection, training, and use.
- In terms of strengthening security management measures.Jitterbug establishes "red and blue confrontation" mechanism, simulating attack cases, fixing potential security vulnerabilities, and optimizing the model's ability to identify false information.
- In terms of implementing content marking requirements.Ali, fast, rare Yu and other key platforms to actively promote metadata implicit identification landing implementation.
- In terms of preventing security risks in key areas, Little Red Book inputs data from specialized areas in the post-training phase of the model to improve the model's ability to understand problems in key areas such as healthcare, finance, and minors.
As a next step, the Central Internet Information Office will focus on seven types of prominent issues such as AI rumor mongering and vulgar content.In addition, we will carry out the second phase of the special action "Clear and rectify the misuse of AI technology", build a technical monitoring system, form a specification for disposal and punishment, promote the content labeling on schedule, and form a long-term working mechanism, so as to strive to maintain a clear network ecology and promote AI to the good and the good.