At the World Internet Conference's Wuzhen Summit, 360 Digital Security Group unveiled its White Paper on Large Model Security Risks and Governance Strategies. The document underscores a critical concern: as advanced AI technologies are increasingly integrated into key industries, the heightened intelligence of AI systems is accompanied by a proportional escalation in security vulnerabilities. The white paper meticulously identifies five primary security threats to large models, encompassing risks in infrastructure, content authenticity, data and knowledge repositories, intelligent agent operations, and user-end interactions. These threats are not isolated but interwoven, creating a complex web of security challenges.

To counter these risks, the white paper advocates for a two-pronged governance approach: "external security safeguards + intrinsic platform security." The former strategy centers on fortifying large models against external threats, while the latter emphasizes the integration of security measures at every developmental stage of large models. Furthermore, 360 introduces a comprehensive security solution that spans the entire lifecycle of large models. This is achieved through the deployment of seven core products, forming a multi-tiered defense mechanism that has already been successfully implemented across diverse sectors.

Looking ahead, 360 is committed to fostering collaboration with industry stakeholders to advance the development of security standards and facilitate technology sharing. The goal is to collectively construct a secure and reliable AI ecosystem that inspires trust and confidence.