Qi An Xin Technology (SH:688561) Zhong Guo Xin Wen Wang·2025-04-29 14:14Core Viewpoint - The application of large models in artificial intelligence (AI) presents both opportunities and challenges, necessitating a focus on security measures to ensure safe deployment and operation [1][3]. Group 1: Key Challenges in AI Model Application - The first challenge is the security of "small data," which is crucial in the digital economy, as its loss can have severe consequences [3]. - The second challenge involves the risk of operational disruptions due to the unregulated deployment of large models, which can lead to a chain reaction of business failures [3]. - The third challenge is the over-reliance on AI, which can result in difficulties in distinguishing between good and bad outcomes, potentially causing significant repercussions if model decisions are flawed [3]. Group 2: Five Key Security Measures - The first measure is to ensure secure access by creating a "red zone" for large model applications, which involves multi-dimensional boundary isolation for data, computing power, and management [4]. - The second measure emphasizes strong control over permissions, utilizing technologies like bastion hosts and zero trust to enhance security for special personnel and terminals involved in development and training [4]. - The third measure focuses on stringent content governance throughout the lifecycle of large models, ensuring core data is monitored and controlled effectively [4]. - The fourth measure involves practical assessments of large model applications to identify and mitigate new security risks, ensuring comprehensive evaluation of various safety aspects [4]. - The fifth measure is to establish a closed-loop operation for real-time monitoring and emergency response to security threats, integrating AI capabilities into security operations for effective detection and interception of attacks [5].