Qi An Xin's Qi Xiangdong: Security Is a Must-Answer Question as AI Agents Reshape Software

Qi, who is also a member of the National Committee of the Chinese People's Political Consultative Conference and vice chairman of the All-China Federation of Industry and Commerce, delivered a keynote address titled "Intelligent Agents Reshaping Software: Raise Efficiency, Keep Risk Under Control," according to Leifeng Net.

He pointed to the Ministry of Industry and Information Technology's "AI Plus Software" special action implementation plan, issued on Sept. 11, which calls for cultivating a group of high-level intelligent programming tools and intelligent development platforms by 2028, extending their use to 20,000 software enterprises above a designated size, and building 100 benchmark agent-based software applications in key industries. By 2030, the plan states, key software should be fully upgraded with intelligent capabilities. The deeper agents penetrate software, Qi said, the more security becomes a compulsory question.

Qi described digital employees as the strongest current engine for raising productivity and predicted they will become the backbone of enterprises. They can write code across the full chain of software design, development, testing, launch and operations; work in offices handling data and producing reference material for decisions; and help manage supply chains by checking orders, taking inventory and warning of swings in supply and demand.

In cybersecurity, he said, digital employees have comparable potential: a small number can carry out routine inspections, and when attack alerts surge, adding computing power can expand defensive capacity almost instantly. Qi An Xin's security operations digital employee reaches a 100 percent alert triage rate, he said, while its threat analysis digital expert covers 32 high-frequency threat scenarios including vulnerability assessment, malware sample analysis and APT tracing. For research, development and office work, the company has built the AI-native development platform QAgent Fabric and the LanMate AI work execution platform, embedding security into the default settings of AI programming.

Qi listed five risks that he said must be guarded against as agents reshape software: security rules failing, with flawed software going live; software emerging in scattered form, leaving asset inventories unclear; permission boundaries breaking down and exposing sensitive data; development materials and agent permissions becoming new attack entry points; and systems turning into black boxes that are hard to restore and hard to hold accountable.

His proposed principle is to keep agents "inside a circle" and manage their "hands and feet" so they can be used safely. That translates into five circles — deployment, connection, permission, data and audit — with an agent security management platform acting as the brain and an AI security gateway as the hub, linked to skills detection, zero trust and an AI security operations center. Together they form a closed loop running from discovery and assessment through location and handling to optimization, turning the efficiency gains of agent-driven software into secure, trusted productivity, he said.

Qi added that Qi An Xin, which describes itself as a leading next-generation cybersecurity company, will work with technology companies to build a protection ecosystem for the intelligent era.