|
Article 50
|
(Principles of security and stability)
- When AI is used during model building and validation stages (including pre-trainings and optimization trainings), an organization shall pay attention to the security and take effective measures, including but not limited to data quality processing, model validation and monitoring, in selection of models or algorithms and other relevant tools, to enhance training quality and prevent generation of improper information with the purpose of increasing accuracy and reliability of contents exported or generated by AI systems.
- An organization shall comply with information security regulations and create appropriate information security protection or management and control measures to prevent security threats and attacks, such as hacker attack and malware, and continue to monitor operation results to ensure the security of AI systems.
|