1. Background of the promulgation
Accordingto the Interim Measures for the Management of Generative Artificial Intelligence Services, providing generative AI services with public opinion attributes or social mobilization capabilities to the domestic public should carry out security assessment and handle algorithm filing in accordance with the law to ensure that the service operation is safe and controllable, and the content is legal and compliant.
2. Scope of application
includes institutions and enterprises that self-research, fine-tune, custom training, or provide large model services to the public through APIs, especially platforms with greater influence and dissemination capabilities.
3. Filing and evaluation material
- security assessment report: covering the legality of training data, model security, risk prevention and control measures, emergency response plans, etc., can be prepared with reference to the "Basic Requirements for Generative Artificial Intelligence Service Security".
- Algorithm filing table: Contains information such as model name, development unit, algorithm type, training resources, and content review mechanism.
- Auxiliary materials: such as test question sets, interception keyword bases, corpus annotation rules, etc., to support security testing and evaluation.
4. Process requirements
Complete the security assessment before submitting the algorithm filing application. The process includes material preparation, submission, review by local internet information departments, and publicity of filing information. The processing cycle is generally 3 to 6 months.
5. Follow-up obligations
After the service is launched, the filing number should be publicized in a prominent position or in the product description, and content security management should be continuously implemented. Services that fail to fulfill their safety assessment and filing obligations may face removal, rectification or penalties.
See the official website for details:
https://www.cac.gov.cn/2024-04/02/c_1713729983803145.htm
