Zscaler Expands Zero Trust SASE to Secure AI Agents and Enhance Cybersecurity

Zero trust architecture has spent years moving from a security philosophy to an operating model. This week matters because the conversation is no longer just about “never trust, always verify” for humans and laptops—it’s increasingly about securing machine-to-machine communications as AI agents proliferate across enterprise workflows. In the June 16–23, 2026 window, the most concrete, time-bound development came from Zscaler’s June 16 announcement expanding its Zero Trust Secure Access Service Edge (SASE) approach for what it frames as the “AI era.” [1]

The underlying tension is straightforward: enterprises are being pushed to adopt AI-driven automation while simultaneously dealing with unmanaged devices, browser-to-workload access patterns, and adversaries who can also use AI to scale attacks. In that environment, zero trust becomes less about a perimeter replacement and more about a consistent enforcement layer across identities, devices, sessions, and workloads—especially when the “actor” might be an AI agent rather than a person.

Zscaler’s messaging this week also reinforces a broader industry narrative: the security platform that can observe and control the most transactions—and apply policy consistently across them—argues it can better protect emerging AI-agent communication patterns. Zscaler executives and partners pointed to the company’s telemetry and customer footprint as a differentiator for securing AI agents, framing zero trust as foundational for deploying them safely. [2] [3]

What follows is a grounded look at what was announced, why it matters to zero trust architecture, and what engineering leaders should take away as they plan for AI-enabled enterprise operations.

What happened this week: Zscaler expands Zero Trust SASE for “the AI era”

On June 16, 2026, Zscaler announced an expansion of its Zero Trust SASE solution, positioning the update as a response to unmanaged devices and AI-driven attacks. [1] The company introduced the ZAgent Framework and described innovations intended to secure communications “from browser to workload” using a cloud-native architecture. [1]

Two details in the announcement are particularly relevant to zero trust architecture as practiced (not just theorized). First, the emphasis on browser-to-workload security highlights how much enterprise access has shifted away from traditional network paths. If the browser is the new “client,” then enforcing policy at that interaction point becomes central to zero trust outcomes—especially when devices are unmanaged or partially managed. [1]

Second, Zscaler framed the scale of its enforcement and inspection as a core capability, stating it can secure over 750 billion daily transactions. [1] While transaction volume alone doesn’t prove security efficacy, it does signal the company’s strategy: use large-scale telemetry and cloud-native enforcement to apply zero trust controls broadly and consistently.

Although the date range for this weekly insight is June 16–23, the week’s context also includes Zscaler leadership and partner commentary from June 11–12 that directly ties zero trust platforms to AI agent operations. Zscaler CEO Jay Chaudhry argued that zero trust security platforms are best equipped to protect the communications needed for AI agents, and partners echoed that Zscaler’s telemetry and global system for securing communications between AI agents provides an advantage. [2] [3]

Why it matters: Zero trust is shifting from user access to agent communications

The most important architectural shift implied by this week’s coverage is that “who/what needs access” is expanding. Zero trust programs historically focused on users, endpoints, and applications. The Zscaler narrative reframes the problem around communications—particularly the communications that AI agents require to operate across tools, data sources, and workloads. [2]

That matters because AI agents can multiply the number of access decisions an organization must make. If an agent is acting on behalf of a user or a workflow, the security system must still enforce least privilege and verify context continuously. The week’s reporting positions zero trust as the “foundation” for deploying AI agents, precisely because it is designed to authenticate and authorize communications rather than assume trust based on network location. [2]

The unmanaged device angle is equally consequential. Unmanaged endpoints are a persistent gap in many zero trust rollouts: policy can be defined, but enforcement becomes inconsistent when devices fall outside management tooling. Zscaler’s June 16 update explicitly calls out unmanaged devices as part of the challenge set it aims to address, tying that to a cloud-native approach that secures browser-to-workload communications. [1]

Finally, the “AI-driven attacks” framing is a reminder that adversaries also benefit from automation. If attacks scale, then verification and policy enforcement must scale too. Zscaler’s emphasis on securing hundreds of billions of daily transactions is effectively an argument that zero trust controls must operate at massive throughput to remain practical in AI-accelerated environments. [1]

Expert take (from the week’s reporting): Telemetry and scale as a zero trust differentiator

This week’s sources converge on a specific thesis: the platform with the most visibility into real-world communications can better secure AI-agent interactions. Zscaler’s CEO emphasized that zero trust platforms are best suited to protect the communications required for AI agents, and he attributed Zscaler’s lead to its customer base and “massive data telemetry.” [2]

Partner executives reinforced the same point, describing an “incredible advantage” tied to Zscaler’s large customer base, extensive telemetry, and a global system for securing communications between AI agents. [3] Taken together, the argument is that telemetry is not just a monitoring asset—it becomes an input to policy enforcement and threat detection at the communication layer.

From a zero trust architecture perspective, this is a notable emphasis. Zero trust is often described in terms of identity, device posture, and micro-segmentation. The week’s framing shifts attention to the operational substrate: the continuous stream of transactions and sessions that must be evaluated and controlled. [1] [2] [3]

It’s also a reminder that “zero trust” in the market is increasingly packaged as a platform story—SASE plus agent frameworks plus cloud-native enforcement—rather than a set of discrete controls. Zscaler’s June 16 announcement explicitly positions its Zero Trust SASE evolution as a response to the AI era, implying that the next phase of zero trust competition will be about who can secure new communication patterns (browser-to-workload, agent-to-agent) without breaking performance or usability. [1]

Real-world impact: What security and platform teams should watch right now

For engineering and security leaders, the practical impact of this week’s developments is less about adopting a specific vendor feature and more about validating whether your zero trust architecture can handle emerging access patterns.

First, browser-to-workload security is being treated as a first-class problem. If your organization’s critical workflows increasingly run through browsers and SaaS interfaces, then the browser becomes a key enforcement point for policy and inspection—especially when devices are unmanaged. Zscaler’s announcement explicitly targets this path. [1]

Second, AI agents introduce a communications-centric security requirement. If agents are going to operate across systems, the security model must protect the communications they rely on. Zscaler’s CEO explicitly framed zero trust as foundational for deploying AI agents, which should prompt teams to ask: do we have consistent authentication, authorization, and policy enforcement for non-human actors and their sessions? [2]

Third, scale and telemetry are being positioned as strategic assets. Zscaler’s claim of securing over 750 billion daily transactions underscores the operational reality: zero trust enforcement must be high-throughput and resilient to be viable at enterprise scale. [1] Whether or not your organization operates at that magnitude, the direction is clear—AI-driven workflows can increase transaction volume and complexity, and security controls must keep up.

Finally, partner commentary suggests the ecosystem is aligning around the idea that securing AI-agent communications will be a major near-term battleground. [3] That should influence roadmap priorities: inventory where agents will run, what they will access, and how communications will be authenticated and governed.

Analysis & Implications: Zero trust is becoming the control plane for AI-era enterprise operations

This week’s limited but pointed set of developments suggests a maturation of zero trust architecture from “access modernization” to “communications governance at scale.” Zscaler’s June 16 expansion of Zero Trust SASE—highlighting the ZAgent Framework and browser-to-workload security—signals that vendors see the browser and cloud workloads as the primary arena for enforcement, particularly when endpoints are unmanaged. [1]

The AI angle changes the stakes. In the reporting, Zscaler leadership argues that zero trust is the real foundation for deploying AI agents because it protects the communications those agents require. [2] That framing implies a shift in what “identity” means in zero trust: not only human identities and device identities, but also agent identities and the sessions they initiate. Even without additional technical detail in the sources, the architectural implication is clear: if AI agents become routine actors in enterprise systems, then zero trust controls must apply to them as consistently as they apply to employees.

The emphasis on telemetry and transaction volume is also telling. Zscaler’s claim of securing over 750 billion daily transactions is presented as evidence of scalability and readiness for AI-era threats. [1] Separately, both the CEO and partners cite massive telemetry and a global system for securing communications between AI agents as a competitive advantage. [2] [3] In zero trust terms, this suggests that policy enforcement and threat detection are increasingly data-driven and platform-centric: the more communications you can observe and mediate, the more confidently you can enforce least privilege and detect anomalies.

For practitioners, the implication is that zero trust programs may need to be evaluated less by whether they have “implemented ZTNA” and more by whether they have built a durable control plane for communications across browsers, workloads, and automated actors. Zscaler’s messaging ties unmanaged devices, AI-driven attacks, and agent communications into one narrative: the enterprise edge is now a set of sessions and transactions, not a network boundary. [1] [2]

This week doesn’t provide a broad industry survey, but it does provide a crisp signal: zero trust architecture is being marketed—and arguably engineered—as the enabling layer for AI adoption, not merely the security tax that comes with it. [1] [2] [3]

Conclusion: Zero trust’s next test is whether it can secure “non-human work”

The June 16–23, 2026 window offered a narrow set of zero trust-specific news, but it was directionally significant. Zscaler’s expansion of Zero Trust SASE, including the ZAgent Framework and a focus on securing browser-to-workload communications, is positioned as a response to unmanaged devices and AI-driven attacks. [1] In parallel, Zscaler leadership and partners argued that zero trust platforms are foundational for deploying AI agents safely, emphasizing the role of telemetry and large-scale communications security. [2] [3]

The takeaway for Enginerds readers is not that every organization must mirror one vendor’s architecture, but that zero trust success criteria are changing. If AI agents become common, the “unit of security” shifts further toward communications: sessions, transactions, and policy decisions made continuously at scale. This week’s reporting suggests that vendors will compete on who can enforce those decisions most consistently across browsers, workloads, and automated actors—without sacrificing performance.

If your zero trust roadmap still assumes a mostly human-driven access model, this is the week to revisit that assumption. The next phase of zero trust architecture may be defined by how well it secures non-human work.

References

[1] Zscaler Redefines Zero Trust SASE for the AI Era — CIO&Leader, June 16, 2026, https://www.cioandleader.com/zscaler-redefines-zero-trust-sase-for-the-ai-era/?utm_source=openai
[2] Zscaler CEO On Why Zero Trust Is The Real ‘Foundation’ For Deploying AI Agents — CRN, June 11, 2026, https://www.crn.com/news/security?utm_source=openai
[3] Zscaler Has ‘Incredible Advantage’ For Securing AI Agent Boom: Partners — CRN, June 12, 2026, https://www.crn.com/news/security?utm_source=openai