Zscaler Adds Zero-Trust Controls for Agentic AI in Enterprise Networks

Zscaler announced new product capabilities for securing agentic AI across its Zero Trust Exchange platform, including AI Broker with Agent Registry, Endpoint AI Security, AI Access Graph, and expanded AI Protect controls.

Zscaler targets agent-to-agent security

In its announcement, Zscaler said the new capabilities are intended to help enterprises manage the security and governance risks created by AI agents that can access applications, exchange data, and act on behalf of users or systems. The company framed the launch around “agentic AI,” a term commonly used for AI systems that can perform multi-step tasks with some degree of autonomy.

According to Zscaler, AI Broker with Agent Registry is designed to provide visibility and control over enterprise AI agents. The company said the feature supports governance for agent communications, including interactions that use agent-to-agent protocols and the Model Context Protocol.

SiliconANGLE reported that Zscaler introduced AI Broker at Zenith Live 2026 and described it as a way to secure communications involving AI agents. The same report said Zscaler is addressing both A2A communications and MCP-based connections, two areas that have become more relevant as companies connect AI systems to tools, data sources, and enterprise workflows.

Endpoint and access controls are part of the rollout

Zscaler also announced Endpoint AI Security, which it said is focused on AI-related risks at the device layer. In the company’s description, the product is meant to help organizations discover and control AI activity on endpoints, including threats or misuse that could arise when AI tools operate on employee devices.

Network World reported that Endpoint AI Security was one of several components in Zscaler’s zero-trust platform update for agentic AI. The publication also named AI Broker, AI Access Graph, and AI Protect expansions as part of the launch.

AI Access Graph is another major piece of the announcement. Zscaler said the feature is intended to map relationships among users, AI agents, applications, and data access paths. That mapping is meant to help security teams understand which agents can reach which resources, and where excessive or risky access may exist.

The company also described new AI Protect capabilities for data and governance controls. Zscaler’s announcement said these expansions are intended to help enterprises secure sensitive data as employees and AI agents use AI services.

The broader issue is enterprise AI governance

The launch reflects a broader shift in enterprise security: AI systems are increasingly being treated not only as applications, but also as actors that may need identities, permissions, monitoring, and policy enforcement. Zscaler’s materials emphasize a zero-trust approach, meaning access should be evaluated continuously rather than assumed based on network location or prior authentication.

That framing is consistent across the third-party coverage. SiliconANGLE reported that Zscaler’s new tools focus on securing AI agents and their communications, while Network World characterized the announcement as a zero-trust platform for agentic AI.

For enterprise buyers, the practical question will be how these controls integrate with existing identity, endpoint, network, and data security systems. Zscaler’s announcement positions the features as part of its Zero Trust Exchange platform, but the company’s claims will need to be evaluated in deployment, especially in organizations already using multiple AI platforms and security vendors.

The announcement also shows how quickly security vendors are adapting their products to AI agents. As more companies experiment with agents that can query systems, trigger actions, and pass information between tools, the security model becomes more complex than monitoring employee use of chatbots. Zscaler is arguing that agent discovery, communication control, endpoint visibility, access mapping, and data protection should be managed together under a zero-trust architecture.

Zscaler did not present the announcement as a single standalone product only. Instead, its press release describes a set of Zero Trust Exchange innovations aimed at securing enterprise agentic AI across access, data, endpoint, and governance layers.