Venustech Group Inc.'s proposal for "Key Technologies and Applications for Non-Human Identity Security Governance in AI Agents" has been officially included in the 2026 Haidian District Association for Science and Technology (HAST) Golden Bridge Project list. The project emerged victorious from a highly competitive field, undergoing multiple rounds of rigorous screening and professional evaluation to become the sole cybersecurity industry entry selected this year.
For the first time, the final assessment was led by an academician, joined by distinguished experts from universities, research institutes, national laboratories, and new R&D institutions. The panel conducted on-site defense sessions and provided face-to-face guidance to the shortlisted teams. This recognition underscores Venustech's forward-looking strategic positioning and deep technical expertise in the field of non-human identity security for intelligent agents.
As AI technology rapidly advances, the interaction patterns among agents, between agents and humans, and between agents and business systems have grown increasingly complex. Establishing a framework where agents possess clear identities, authorization, auditability, and accountability has become a foundational requirement for scaling AI deployment. This imperative aligns seamlessly with the principles championed by Venustech's "Tianguan·Zero Trust+" initiative, which emphasizes trusted identity, dynamic authorization, continuous auditing, and a closed-loop accountability framework.
To address these challenges, the project leverages the core philosophy of "Tianguan·Zero Trust+" and builds upon the MASB Non-Human Identity Security Control System. It extends the "Identity as Security" management paradigm to encompass intelligent agents, tool invocations, and non-human identity governance. By focusing on trusted identity, least privilege access, dynamic policy enforcement, and audit-based accountability, the initiative aims to transform agent operating environments from "anonymous invocation" to "verified and controlled" operations.
Throughout the entire lifecycle of an agent—covering onboarding, operation, invocation, and decommissioning—the MASB system applies its practical expertise in identity trust, dynamic access control, and end-to-end trusted access. It delivers essential functions including identity recognition, credential management, task-level authorization, tool invocation control, behavior auditing, and risk response. These capabilities propel agents from mere functionality toward comprehensive management, control, and trustworthiness.
Specifically, the system offers unified identity management for agents, encompassing registration upon onboarding, identity issuance, status updates, and revocation upon exit. It implements task-level authorization with least privilege controls to mitigate excessive authorization risks during long-running operations. Dual-identity authentication and responsibility tracing enable cross-validation between human users and agents. Semantic-level behavior auditing and AI-native attack detection effectively identify threats such as MCP poisoning and tool misuse.
Addressing the frontiers of AI security and practical industry demands, Venustech will continue advancing the research and deployment of key technologies for non-human identity security in intelligent agents. The company remains committed to refining its capabilities in identity management, dynamic permissions, behavioral auditing, risk detection, and accountability loops, thereby contributing to Haidian District's AI innovation ecosystem and the growth of China's AI security industry.