Zscaler Unveils AI Broker for Enhanced Agentic AI Security at Zenith Live 2026

Zscaler has announced a suite of new security innovations at its Zenith Live 2026 conference, with a strong emphasis on securing agentic AI systems. The company's latest offerings, including AI Broker, AI Access Graph, and the ZAgent Framework, are designed to provide comprehensive governance over how AI agents communicate, access data, and operate across enterprise networks. This strategic push positions Zscaler at the forefront of a rapidly evolving market where major security vendors are all vying to establish the definitive control plane for AI agent security. A cornerstone of Zscaler's new security paradigm is the AI Broker. This system is specifically engineered to secure agent-to-agent (A2A) and Model Context Protocol (MCP) communications. It incorporates an integrated Agent Registry, which is crucial for tracking and managing the permissions granted to individual AI agents. By doing so, AI Broker can apply fine-grained access controls, ensuring that agents only interact with authorized resources and other agents. This addresses a critical gap in current enterprise security, as the agent-to-agent communication layer is largely uninstrumented, allowing agents to invoke tools and retrieve data from sources that may bypass existing access control policies. Furthermore, Zscaler introduced Endpoint AI Security, which extends threat detection capabilities to cover browser extensions, plugins, and locally running AI tools on employee devices. This is vital as AI agents become more prevalent on user endpoints, potentially introducing new attack vectors. The company's AI Access Graph, built upon the recent acquisition of Symmetry Systems, provides a crucial visibility layer by mapping identity and data lineage across the entire enterprise. This foundational insight is essential for the effective functioning of the other security controls. The timing of these announcements coincides with a broader industry effort to define runtime governance for AI agents, as evidenced by the recent launch of the Agent Control Standard (ACS). While ACS offers a vendor-agnostic approach, Zscaler's proprietary solution presents a trade-off for enterprises. A significant challenge remains in the adoption and implementation of these security frameworks: the sheer proliferation of AI agents often outpaces IT's ability to register and classify them. Without accurate cataloging, any governance framework faces friction. Zscaler's innovations aim to provide the necessary tools to manage this complexity, ensuring that as AI agents become integral to business operations, their security and compliance can be maintained with the same rigor as other mission-critical enterprise workloads.