ARMOR In Action with Illumio: Containing Breaches Across AI Infrastructure

As AI evolves from pilot projects into business-critical systems, security becomes a defining factor in successful AI deployments. To help organizations address this challenge, WWT developed ARMOR (AI Readiness Model for Operational Resilience), a vendor-agnostic AI security framework. Built through collaboration with NVIDIA and strengthened by real-world engagement with the Texas A&M University System, ARMOR provides security leaders with a practical framework for assessing and managing AI risk. It organizes AI security into seven domains: Governance, Risk and Compliance, Model Protection, Secure AI Operations, Infrastructure Security, Identity Security, Data Protection, and Secure Software Development Lifecycle. Together, these domains help organizations build cyber-resilient AI environments

ARMOR's seven domains focus on key elements of securing AI systems and related processes, helping organizations understand security control points inside and between these systems, and their interactions with the surrounding environment. This is key to understanding how AI security mechanisms mesh with traditional security solutions and is vital to identifying present gaps. ARMOR also accommodates organizations at different maturity levels. ARMOR is a real-world-tested AI security framework that was shaped through two-way collaboration with a lighthouse customer, Texas A&M University, who served as the first testing ground.

Illumio brings a distinctive perspective to the ARMOR framework. Agentic AI is increasing the speed and scale of attacks, and when AI agents move through enterprise infrastructure at machine speed, the ability to see and control lateral movement becomes a defining control. Illumio is built for exactly that problem. Named a Leader and a Customer Favorite in The Forrester Wave™: Microsegmentation Solutions, Q3 2026, it focuses on breach containment: assume attackers will get in, so limit how far they can move. Illumio's security portfolio, made up of Illumio Segmentation and Illumio Insights and built on what Illumio calls its AI security graph, applies one label-based policy model across data centers, clouds, containers, endpoints and OT. It enforces through its agent on the workloads where one can run, and through cloud security groups, switches, load balancers, firewalls and DPUs where it can't. Illumio positions its role in agentic AI as containing threats at the infrastructure layer, working alongside prompt, model and agent governance controls rather than replacing them. In the sections ahead, we'll show how Illumio's solutions map to the ARMOR domains, with particular strength in Infrastructure Security, Secure AI Operations and the governance and cyber-resilience controls that show whether containment is working.

Mapping Illumio's portfolio to ARMOR domains

Illumio's portfolio is organized around a continuous loop: see how everything connects, define least-privilege policy and contain attacks quickly once they are detected. Illumio Insights provides the visibility and detection layer, Illumio Segmentation provides enforcement across data center, cloud, container, endpoint and OT environments, and both share the same AI security graph. This design aligns well to ARMOR's domain structure. Illumio's strongest coverage sits in Infrastructure Security (Illumio Segmentation) and Secure AI Operations (Illumio Insights, Insights Agent and SOC integrations), and in GRC, where Network Posture measures containment against compliance standards, Segmentation keeps a read-only audit record, and the Insights Known Public LLMs view shows which public LLM services workloads and accounts are connecting to.

Across the seven ARMOR domains, Illumio's alignment can be summarized as follows:

With the portfolio mapped, the following sections describe how each Illumio solution contributes to the ARMOR domains it primarily supports and the domains it secondarily reinforces.

Illumio Segmentation for Data Centers

ARMOR Domains: Infrastructure Security (Primary); Governance, Risk and Compliance, Identity Security, Data Security (Overlap)

AI training clusters, inference services, model repositories and the data stores that feed them are high-value targets, and once an attacker is inside, lateral movement is how one compromised host becomes an AI-wide incident. Illumio Segmentation ringfences these applications, enforcing least-privilege communication between workloads from a single policy engine, independent of the underlying network. It is the foundation of Illumio's Infrastructure Security coverage. Notable capabilities include:

• Dependency Mapping and Policy Modeling: Maps connections between servers, virtual machines and containers from flow telemetry reported by the Virtual Enforcement Node (VEN) and ingested sources. The Policy Generator recommends rules from observed traffic, and Draft View shows how known flows would be handled before a policy is enforced.
• Label-Based, Default-Deny Policy: Builds policy from workload labels rather than IP addresses. In full enforcement, anything not explicitly allowed is blocked, inbound and outbound, and workloads can run in visibility-only or selective enforcement modes first. New workloads that receive the right labels, for example through pairing profiles, inherit the matching rules.
• Host-Based Enforcement with Tamper Detection: The VEN programs the host's own firewall (iptables, nftables, Windows Filtering Platform and others), detects changes to those rules and restores the last-known-good policy. Optional tamper protection requires a maintenance token to stop, suspend or uninstall it.
• Coverage Across the Hybrid Estate: The VEN enforces on Windows, Linux, Solaris and AIX. Where an agent can't be installed, the Network Enforcement Node translates the same label-based policy into rules on Cisco and Arista switches, F5 BIG-IP, AVI and Citrix ADC load balancers, and IBM i systems running Precisely Assure Security, with z/OS mainframe coverage through BMC.
• Identity-Based Policy and Encryption in Transit: Adaptive User Segmentation writes rules from Active Directory and, where enabled by Illumio, Microsoft Entra ID user groups. SecureConnect encrypts traffic between workloads with IPsec, as defined by policy.
• Audit-Ready Enforcement Record: Every policy change and administrative action is recorded as a read-only audit event that can't be disabled. Events export to SIEM tools, giving compliance teams evidence that segmentation controls are in place.

Illumio Segmentation for Cloud & Containers

ARMOR Domains: Infrastructure Security (Primary); Governance, Risk and Compliance, Secure Software Development Lifecycle (Overlap)

AI applications increasingly run as containerized services across multiple clouds, where east-west traffic is hard to see and controls differ from account to account. Illumio uses a shared label model across data center, cloud and container policy, with enforcement points and label types suited to each environment, including the CI/CD systems that build AI applications. Core capabilities include:

• Multi-Cloud Visibility and Agentless Enforcement: Provides one view of traffic and risk across AWS, Microsoft Azure, Google Cloud and Oracle Cloud Infrastructure. Policy is enforced through AWS security groups and NACLs, Azure NSGs and Azure Firewall, and Google Cloud firewall rules, while OCI enforcement is in beta.
• Container Segmentation: The container agent enforces policy at the pod level in Kubernetes and OpenShift. An agentless option adds visibility of Kubernetes traffic on managed services such as AKS, GKE and OpenShift OVN-Kubernetes.
• Policy That Follows the Workload: Labels, not IP addresses, drive policy, so a rule written for a production inference tier applies to new instances that carry the same labels as the tier scales.
• Policy Drift Correction: Tamper protection re-applies Illumio-written security group and firewall rules if someone changes them manually, so cloud controls stay as approved.
• CI/CD and Build Pipeline Containment: In full enforcement, VEN-managed build and pipeline nodes are default-deny inbound and outbound, so they reach only explicitly allowed destinations. A breached CI/CD system is limited to what policy allows, which restricts lateral movement into production.

Illumio Segmentation for Endpoints

ARMOR Domains: Infrastructure Security (Primary); Secure AI Operations (Overlap)

User devices remain one of the most common entry points into an environment, and developer and data science workstations can hold credentials and paths into sensitive systems, including AI platforms. Illumio Segmentation applies the same policy model on endpoints at the operating-system level, so the segmentation that protects data center and cloud workloads extends to the devices that reach them, and traffic from a compromised device is limited to what policy allows. Highlighted capabilities include:

• OS-Level Containment: Enforces policy on the device itself, so connections from a compromised laptop to the file server or other workloads next door are blocked unless policy allows them.
• Quarantine: Compromised endpoints can be isolated by applying a quarantine label that moves them into a restrictive policy, triggered manually or automatically from tools such as Splunk, Microsoft Sentinel or Zeek.
• One Policy Engine: Uses the same policy engine that protects cloud workloads and data centers, so device-to-workload access follows the same least-privilege model.

Illumio Insights

ARMOR Domains: Secure AI Operations (Primary); Governance, Risk and Compliance, Infrastructure Security, Data Security (Overlap)

Securing AI in production means knowing, continuously, how AI workloads and the people and services around them communicate, and spotting the moment that behavior changes. Illumio Insights is the visibility, observability and detection layer of the Illumio platform. Built on the AI security graph, which maps observed communication across workloads, devices and the internet, it observes east-west traffic across hybrid and multi-cloud environments, surfaces lateral movement, shows which workloads are reachable from vulnerable ones and connects detection to containment. Notable capabilities include:

• Known Public LLMs Dashboard: Shows which publicly accessible LLM services an organization's resources are accessing and exchanging data with, and which resources and accounts are involved, giving Shadow AI governance programs visibility based on observed traffic.
• AI Agent Process Detection: Detects AI agent processes and shows them in the map alongside other workloads, so teams can see what agents connect to.
• Lateral Movement and Anomaly Detection: Flags lateral movement, compromised workloads and high-risk traffic, and the Anomaly Detection dashboard compares behavior with a 30-day baseline to surface deviations.
• Data Exfiltration Detection: Tracks unusual outbound activity and flags the workloads responsible, so teams can investigate and contain them.
• Cloud Quarantine: From an Insights finding, teams can quarantine cloud virtual machines on AWS, Microsoft Azure and Google Cloud in one click, enforced through the provider's security groups or firewall, with controlled administrative access kept open.
• Agentless Hybrid Visibility: Ingests Check Point and Fortinet firewall logs into Insights to map data center traffic without agents on the workloads.

Illumio Insights Agent

ARMOR Domains: Secure AI Operations (Primary); Governance, Risk and Compliance (Overlap)

Security teams already face more alerts than they can triage, and AI-driven attacks compress the time available to respond. Insights Agent is an AI-powered, role-aware guide built into Illumio Insights that prioritizes what matters on each workload and turns findings into clear next steps. It also reaches analysts through Microsoft's AI ecosystem: Illumio Insights was a launch partner in the Microsoft Marketplace AI apps and agents category and is also listed in the Microsoft Security Store. Capabilities include:

• Persona-Based Guidance: Tailors findings and alerts for four roles: threat hunter, incident responder, data security and compliance monitor.
• Investigative Analysis: Analyzes workloads, policies and traffic flows and ranks recommendations by severity.
• AI-Driven Response Plans: Guides users through prioritized, step-by-step remediation.
• MITRE ATT&CK Mapping: Maps anomalous behavior to MITRE ATT&CK and explains what each finding means, why it matters and how to respond.
• Guided Containment: Connects recommendations to containment actions, including one-click quarantine of cloud virtual machines.

Illumio Network Posture

ARMOR Domains: Governance, Risk and Compliance (Primary); Secure AI Operations, Infrastructure Security (Overlap)

ARMOR treats governance as something organizations must be able to prove, not just document. Network Posture, a capability within Illumio Insights released in 2026, measures lateral movement risk by validating observed flows against policy intent and guardrails aligned to industry standards, giving teams a current view of whether containment controls are working. Core capabilities include:

• Real-Time Lateral Movement Risk Score: Validates live network flows, not just static rules, to score where lateral movement risk exists, including exposures that are not yet being exploited.
• Standards-Aligned Scoring: Scores live traffic against guardrails aligned to standards including PCI DSS, SOC 2, CIS, HIPAA, ISO/IEC 27001, NERC CIP, NIST SP 800-82 and IEC 62443, as well as Illumio best practices and custom guardrails. This is guardrail-based alignment, not a compliance certification, and Illumio also cites alignment with DORA, GDPR and NIST CSF.
• Prioritized Remediation: Ranks guardrail rules by number of violations and, through Vulnerability Exposure, ranks application groups by how many workloads are reachable from vulnerable ones, so teams know where to tighten segmentation first.

Illumio Ecosystem Integrations

ARMOR Domains: Secure AI Operations (Primary); Governance, Risk and Compliance, Infrastructure Security (Overlap)

No single tool secures an AI environment, and ARMOR emphasizes how controls mesh with the rest of the security stack. Illumio feeds its traffic, risk and containment context into the SOC, IT operations and exposure-management tools organizations already run, and exposes most platform functions through REST APIs for automation. Highlighted integrations include:

• Microsoft Security: The Microsoft Sentinel solution brings Illumio audit, flow and workload data into Sentinel and includes playbooks that quarantine workloads through the Illumio API. A Sentinel data lake connector brings Illumio data into the Sentinel data lake, and an Illumio Security Copilot agent lets analysts investigate data lake sources in natural language.
• SIEM and SOAR Containment: The Splunk app and add-on map Illumio data to CIM, and an Enterprise Security adaptive response quarantines workloads from Splunk. Illumio also provides an IBM QRadar app and a Cortex XSOAR and XSIAM content pack for rapid ransomware containment.
• Perimeter Plus Containment: Under their June 2026 partnership expansion, aimed at frontier AI-powered attacks, Check Point and Illumio say customers can align Check Point firewall policy with Illumio's workload model.
• OT and Exposure Context: Enriches risk analysis with OT and IoT device intelligence from Armis and Claroty, and with cloud context from Wiz.
• IT Operations: A certified ServiceNow CMDB app syncs CMDB fields into Illumio labels, and Illumio works with ServiceNow ITSM.
• Agentic AI Guardrails with Netskope: Illumio and Netskope describe a joint approach in which Netskope inspects AI agent and MCP traffic, while Illumio provides workload visibility, least-privilege enforcement and dynamic quarantine, and can signal Netskope to revoke access.

Illumio also ingests Fortinet and Palo Alto Networks firewall logs, imports Qualys and Tenable vulnerability data, and supports Terraform and Ansible for automation.

Illumio Segmentation on NVIDIA BlueField™ DPUs

ARMOR Domains: Infrastructure Security (Primary)

In performance-sensitive environments, security enforcement can't become the bottleneck. Illumio's work with NVIDIA brings segmentation enforcement onto NVIDIA BlueField™-3 DPUs, so policy is applied at the DPU instead of on the host. Capabilities include:

• DPU-Based Enforcement: Illumio's Network Enforcement Node programs IPv4 port, protocol and IP address rules onto Open vSwitch on BlueField-3, so enforcement happens at the DPU. Illumio says this offloads segmentation from the host CPU to minimize impact on application performance.
• Agentless Workload Discovery: A VEN running on the DPU's Arm cores discovers host workloads without installing an agent on the hosts.
• Unified OT and Host Policy: The same label-based policy applies on the DPU path within OT environments and to and from managed host VMs.

Ready to get started?

At WWT, we're helping organizations operationalize ARMOR with the partners who can deliver against it. Illumio's portfolio brings particular strength in Infrastructure Security, Secure AI Operations and the governance and cyber-resilience controls that show whether containment is working; anchored by Illumio Segmentation for least-privilege enforcement across the workloads that host AI and the wider hybrid estate, Illumio Insights and Insights Agent for detection, AI usage visibility and guided containment, and Network Posture for continuous, standards-aligned measurement of lateral movement risk.

Whether you're standing up a new AI platform or bolstering the security posture of an existing one, the path forward starts with visibility and prioritization. Access the ARMOR dashboard, connect with WWT for an ARMOR briefing and take the next step toward secure, resilient AI built on Illumio's breach containment platform.