DAS-Security(SH:688023) Nan Fang Du Shi Bao·2025-05-09 15:20Core Insights - The integration of AI technology into cybersecurity has reached a new stage of intelligent confrontation, with "AI + Security" being the key to breaking through current challenges [1][3] - Anheng Information launched several new products at the "West Lake Sword AI + New Product Launch Conference," including Hengnao 3.0 and AI-driven security solutions tailored for various core scenarios [1][4] Product Developments - Anheng Information introduced Hengnao 3.0, AiLPHA Intelligent Security Operation Platform, AI + SAAS-XDR, and a one-stop platform for AI-driven DevSecOps, marking significant advancements in multi-modal interaction and connectivity [1][4] - The Hengnao 3.0 platform supports MCP protocols and A2A protocols, enabling the intelligent agent to utilize hundreds of plugins, enhancing task execution control and human-machine collaboration efficiency [4] Security Challenges and Solutions - The industry consensus is that general large models face challenges in security scenarios due to a lack of specialized training data, leading to semantic misjudgments in threat analysis [3] - Anheng Information emphasizes the need for "professional intelligent agents" that understand offensive and defensive logic and can dynamically evolve to meet compliance requirements and respond to new threats [3] Government and Industry Applications - Various local governments are launching AI + government applications, which require tailored AI security solutions to address unique industry characteristics and data types [5] - Anheng Information's CTO highlighted the development of a dual-spiral framework that integrates technology and application, focusing on making security smarter and intelligence safer [5] Broader AI Applications - AI is being widely applied across various fields, including natural language processing, computer vision, and healthcare, with the development of large models entering a commercial application phase [6] - The emergence of "hallucination" phenomena in large models raises concerns about content credibility and fairness, necessitating a multi-dimensional approach to ensure safety in AI applications [6]