Zscaler announces a series of security features for agentic AI. The company is expanding its Zero Trust Exchange platform with solutions that secure autonomous AI agents. This involves how they connect, access data, and run on devices. With this, Zscaler is the first to offer a complete Zero Trust platform for agentic AI. Additionally, new companies are joining Project AI-Guardian. This Zscaler initiative focuses on the adoption and governance of AI agents.
Zscaler announced this during its annual Zenith Live conference. Traditional security tools are built around known human identities, but autonomous AI agents are changing that model. They operate at machine speed, create temporary identities, and perform tasks in ways that conventional tools cannot fully see or control. Zscaler’s 2026 AI Threat Report previously revealed an 83 percent year-over-year increase in AI activity among enterprise customers, while the governance gap is widening.
Tip: Zscaler is accelerating the adoption of zero trust for AI
AI Broker and Endpoint AI Security
Two new products take center stage. Zscaler AI Broker secures communication between AI agents via MCP and A2A brokers. An integrated Agent Registry gives organizations insight into which agents have access to which resources and enables granular access controls. With this, Zscaler addresses risks identified by the OWASP Top 10 for agent-based applications, including uncontrolled agent communication.
Zscaler Endpoint AI Security focuses on a different area: end-user devices. The solution detects and blocks AI-related threats in browsers, plugins, extensions, and local AI tools. These are precisely the layers that traditional endpoint security solutions typically lack.
AI Access Graph and Extensions to AI Protect
A third pillar is Zscaler AI Access Graph, made possible by the recent acquisition of Symmetry Systems. This component maps how identities, applications, and data sources are interconnected within an organization and how data flows in real time. Based on these insights, organizations can enforce policies and reduce unnecessary access.
The announcements build on Zscaler AI Protect, launched in January 2026. That solution is now being expanded in three areas: AI Asset Management, Secure Access to AI, and Secure AI Infrastructure. New features include AI red teaming for MCP servers, a standalone prompt hardening service, and compliance heat maps.
“Traditional security was never designed for millions of autonomous agents that act and reach sensitive data at machine speed,” said Jay Chaudhry, CEO of Zscaler.
Tip: Zscaler acquires AI security firm Symmetry Systems
New Project AI-Guardian partners
At the same time, Zscaler is announcing a new phase of Project AI-Guardian. The cybersecurity company is expanding the initiative from global system integrators (GSIs) to Technology Alliance Partners. Eleven new partners are joining, including AWS, CoreWeave, Databricks, Deep Cogito, Equinix, Glean, Google, OpenAI, and Saviynt. GSI partners Coforge and NTT Data are also participating.
The new partners integrate directly with three core components of the Zscaler platform: the AI Access Graph, AI attack surface and risk modeling, and additional governance and protection capabilities. All AI interactions are monitored inline and in real time via the Zero Trust Exchange. Access is continuously verified, and data is inspected as it moves.
Four security domains
The collaboration focuses on zero-trust controls for AI workflows, comprehensive data protection, continuous visibility and governance, and streamlined deployment. Partners integrate their technologies with Zscaler services so that signals and enforcement can be shared across platforms.
“AI is creating enormous opportunities for organizations, but it is also reshaping the threat across the security and governance landscape,” said Dhawal Sharma, EVP- AI Security and Strategic Initiatives, Zscaler. “Securing AI is an ecosystem effort.”
With the first phase of Project AI-Guardian, launched in May 2026 with GSI partners Cognizant, EY, HCLTech, Infosys, TCS, and Wipro, Zscaler had already laid the groundwork for zero trust in enterprise AI usage.
For users, the expansion means a single, consistent control plane for AI, without the integration burden of separate tools. For Technology Alliance Partners, it offers a platform-native route to expand their AI capabilities in the enterprise.