AI智能体在企业流程自动化、数据价值挖掘及多元场景适配方面展现全能优势，但也面临数据泄露和 模型攻击等安全风险。 上能帮企业自动化处理流程、解锁数据价值，下能适配多元场景降低数字化门槛，如今的AI智能体堪 称数字时代的"全能选手"。但超能力也需配上"安全铠甲"，智能体在跨域调用数据、联动各类系统的同 时，也让数据泄露、模型攻击等风险暗藏其中——安永和上海赛博网络安全产业创新研究院联合推出的 《2025全球可信AI治理与数据安全报告》显示，74%的受访企业对AI应用满心期待，却仍被合规风险揪 着心。 这意味着对数据安全的更高要求，以及还需破解实践中的多重困境。网络安全公司也在提出以AI智能 体为核心的解决方案，通过对接企业静态数据库与动态流量，自动化完成数据资产识别、分类分级与风 险监测，再联动防火墙等现有安全组件实现闭环处置。 从积极的角度来看，安全防护体系的升级与AI智能体的发展正呈现双向赋能态势。这种顾虑催生了"AI For Security"与"Security For AI"的双向奔赴模式：一方面，AI技术赋能安全防护，通过威胁预测、智能 合规、自动化响应等功能提升风险管控效率；另一方面，针对AI特有 ...

Core Insights - The recent State Council document emphasizes the importance of "AI governance capabilities" as a key action area, aiming to enhance network space governance through AI technologies [1] - The launch of the "AI Large Model Ecosystem Construction Cluster" in Zhuhai reflects local efforts to address cybersecurity challenges through development and innovation [1][3] Group 1: AI and Cybersecurity Challenges - The rapid development of AI technologies has introduced new cybersecurity risks, including deepfake technology and data bias, which increase the sophistication of cyberattacks [1] - Common cybersecurity threats include the unauthorized use of large models, the mass production of personalized phishing attacks using generative AI, and risks to data security and AI system integrity [1] Group 2: AI as a Solution for Cybersecurity - AI large models are being explored by many tech and cybersecurity companies in China for their ability to better identify and respond to complex security threats compared to traditional tools [2] - AI solutions include intelligent operation, automated risk management, and predictive models for monitoring cyber threats, with applications already in sectors like finance and electricity [2] Group 3: Policy Support and Investment - Zhuhai's policies include a maximum investment of 500 million yuan in "computing power vouchers" to reduce costs for enterprises using AI, and 100 million yuan in "model vouchers" to alleviate financial pressures on R&D [3] - These financial supports aim to accelerate the iteration and upgrade of various AI large models, thereby establishing a technological foundation for the development of the cybersecurity industry [3] Group 4: Future Directions for AI Development - The next step in the large-scale application of AI models may involve creating collective public products that ensure security, as suggested by experts [3] - Establishing a universal security model as an infrastructure service could help reduce compliance costs for large model enterprises and their downstream applications, promoting a safer and more orderly growth of AI technologies [3]