Itdaily - Zscaler unveils zero trust platform for AI agents

Zscaler unveils zero trust platform for AI agents

White Zscaler logo on a blue LED screen with light streaks above an event booth.

Zscaler is launching a zero trust platform for AI agents, extending the Zero Trust Exchange to meet the security needs of AI agents.

Zscaler unveiled a suite of products to secure AI agents at its Zenith Live 2026 conference. The company claims to have built the first complete zero trust platform for autonomously operating AI agents. The platform extends Zscaler’s Zero Trust Exchange to meet the unique security needs of AI agents. These agents operate at machine speed, create temporary identities, and exercise permissions in ways that traditional tools designed for human users cannot keep up with.

New products for agent security

Zscaler AI Broker secures communication between agents and Model Context Protocol servers. The product includes a registry where organizations can define the access levels for each agent. Zscaler Endpoint AI Security focuses on AI threats on employee devices, including browsers, extensions, and local AI tools.

In addition, the company introduced AI Access Graph, a tool that maps the connections between identities, applications, and data sources. This technology stems from the $175 million acquisition of Symmetry Systems in May.

Expansion of existing solutions

Zscaler is building on Zscaler AI Protect, launched in January. The company is adding functionality to discover embedded AI in SaaS and internet traffic, identify agents and MCP servers in the public cloud, and scan agent codebases for risks. It is also expanding controls for approved AI tools, featuring prompt extraction in more than 250 generative AI applications.

For teams developing AI applications, Zscaler is adding red team exercises for MCP servers, an independent prompt hardening service, and compliance heatmaps.