齐向东：智能体重构软件，安全是必答题

TL;DR - Qi An Xin chairman Qi Xiangdong argues that AI agents can transform software development and enterprise operations, but require security controls spanning deployment, connectivity, permissions, data, and auditing. The company is promoting agent platforms and security tooling designed to provide end-to-end governance.

• AI agents could automate software design, development, testing, operations, office tasks, supply-chain management, and cybersecurity workflows.
• Qi An Xin reports that its security agents support 32 common threat scenarios, including vulnerability assessment, malware analysis, and APT tracing.
• Five highlighted risks are ineffective security rules, unmanaged software assets, excessive permissions and data exposure, compromised development materials or agent privileges, and opaque systems that hinder recovery and accountability.
• The proposed architecture combines an agent security-management platform, AI security gateway, skill inspection, zero-trust controls, and security operations into a continuous detection-to-optimization loop.