Everything your workload can reach, it will.
Kubernetes pods, serverless functions, VMs, databases and AI agents each get clearance for the paths they need and nothing else, enforced in the network with nothing to install.
The cloud is not misconfigured. It was never configured.
Egress control and segmentation are not new. Nobody has rolled them out across the cloud because a bad firewall change breaks production, so workloads keep reaching the internet, databases and each other on paths no firewall governs.
DATA LEAVING ON PATHS THE PERIMETER NEVER SEES
Workload traffic leaves through NAT gateways and route tables, not the perimeter firewall, which watches people and the front door while workloads leave by the back.
AI AGENTS REACHING MORE THAN THEIR JOB NEEDS
Agents that share a security group share a source address, so each one inherits the reach of its most permissive neighbor and nobody can say which agent did it.
KUBERNETES PODS LEAVING THROUGH NODE NAT
Pod egress exits through platform-managed NAT gateways and never crosses a central firewall, and policy tied to addresses breaks when pods move.
Enforced in the network without breaking production
Nothing to install, nothing to detect first, and every step is reversible.
See every workload and every path
The Containment Assessment shows every workload, every path and every AI agent no one authorized, and what a compromise could reach.
Insert with one toggle per VPC
Transparent insertion, with no agents, no network redesign and original addresses preserved end to end.
Watch before you enforce
Start in monitor mode, where the Shadow Rule Set shows every connection the policy would have blocked, then promote one policy at a time. Backout is one action.
Keep it current without a ticket queue
Aviatrix Active Defense keeps clearance current as the estate changes, inside guardrails the team sets.
Two controls and an autonomous defense layer, bought together or in any order
A compromised workload can move two ways, out of the estate or sideways to anything else in it. One control covers each direction and the layer keeps both current.
Aviatrix Active Defense
The autonomous defense layer, and the newest part of Aviatrix Clearance.
Its agents build a model of your estate from the estate itself, with no tagging project and no inventory to import, and draft the policy each application needs. They prove each policy in monitor mode on live traffic, so your team sees exactly what would have been denied before anything is, and keep it current as the estate changes. Every policy they write is standard policy your team can read, change or back out in one action, and they never touch a rule a person wrote.
Available today in free early access for design partners on Aviatrix Enterprise, planned for Aviatrix Cloud.
Ask your account team about the design partner programOutbound Containment
Governs egress. A workload reaches only the outside destinations it is cleared for, named by domain rather than IP, with default deny at every workload and no central firewall in the path.
Lateral Containment
Governs east-west. A workload reaches only the other workloads it is cleared for, by identity, across VPCs and clouds, down to the pod, the function, the database and the AI agent.
AI agent discovery and containment runs inside both controls at the agent grain. Priced per workload, and free to every enterprise through May 1.
Security sets the policy and developers never feel it
Enforcement sits at the VPC boundary where egress actually happens, transparent to the workload, so nothing in the build, the deploy or the application changes.
NOTHING IN THE PIPELINE CHANGES
No agents, no SDK, no code change and no proxy to configure. Original addresses are preserved end to end, and the same policy is orchestrated into security groups and Kubernetes NetworkPolicy where a gateway is not needed.
POLICY AS CODE, OWNED BY SECURITY
Clearance is written once on workload identity and lives in Terraform and Kubernetes custom resources. Domains resolve automatically and tags keep groups current, so a new service ships with its clearance instead of a ticket.
ENFORCE WITHOUT A CHANGE WINDOW
Start in monitor mode, read the Shadow Rule Set, promote one policy at a time and back out in one action. Aviatrix Active Defense keeps clearance current as developers ship, inside guardrails your team sets.
You already have partial containment
A next-generation firewall, a user-access proxy or a posture tool each cover part of the problem, and none of them governs the paths a cloud workload actually uses to leave. Measure your stack and Aviatrix Clearance against the same five testable properties.
| Property | Firewall appliances in the cloud | Posture tools without enforcement | Host-agent segmentation | Aviatrix Clearance |
|---|---|---|---|---|
| Path-complete, every path a workload can use is governed | ✕ No | ✕ No | ✕ No | ✓ Yes |
| Identity-aware at Layer 7, policy on workload identity and protocol, not IP | ✓ Yes | ✕ No | ✕ No | ✓ Yes |
| Detection-independent, holds whether or not anything was detected | ✓ Yes | ✕ No | ✓ Yes | ✓ Yes |
| Compute-model agnostic, reaches VMs, pods, functions and AI agents with nothing installed | ✕ No | ✕ No | ✕ No | ✓ Yes |
| Universally propagated, one policy change lands in every cloud and cluster | ✕ No | ✕ No | ✕ No | ✓ Yes |
| Containment delivered | 2 of 5 | 0 of 5 | 1 of 5 | 5 of 5 |
Every one of them misses outbound. Workload egress leaves through NAT gateways and route tables, never crossing the appliance or the proxy, and a posture tool reports it without stopping it. Outbound Containment closes that path and replaces the NAT gateway spend while it does.
Built on the Aviatrix Containment Platform
Aviatrix Clearance runs on the Aviatrix Containment Platform, built on the Cloud Native Security Fabric, and is the second post-quantum offering on the fabric. The platform sets what every cloud workload may reach before any attack and enforces it in the network, at every workload and in every cloud. The Cloud Native Security Fabric is quantum safe.
Whoever owns the workload has a reason to start
Close the gap without risking production
Monitor first and see exactly what would happen before enforcing anything, free through May 1 and every action reversible.
Start free and enforce when your own numbers say so
The Containment Assessment is free to anyone. Through May 1, 2027, Aviatrix Clearance is free to every enterprise across the whole estate under the Clearance Incident Response program, so you have time to patch. Standard terms return after May 1.
Phase 1, the Self-Score
Five minutes and nothing to deploy, for a provisional Blast Radius and a harvest exposure report on your estate.
Start the Self-ScorePhase 2, the Architecture Read
Read-only and agentless, measured against live runtime flows, sizing exposure in dollars as Reachable Value at Risk with outbound and lateral reported separately.
Clearance Incident Response
Outbound, lateral and intrusion prevention across your whole estate, with no workload cap, from November 1 through May 1. Start in monitor mode, watch the Shadow Rule Set, and enforce when nothing you need is on the list. AWS Marketplace enrollment opens November 1.
Join the programAfter May 1, the Free Tier includes a free Containment Assessment and 15 workloads you can run in monitor or enforce mode, with no purchase and no time limit. Fifteen workloads cover a typical VPC or VNet. Contact Aviatrix for pricing beyond that.
One platform. Three families.
The Aviatrix Containment Platform acts in three places, and Aviatrix Clearance is one of its two offers.
- Zero Trust for Cloud Networks
Harvest and Decrypt Protection
Captured traffic stays unreadable and the encryption algorithm changes with a policy push, the first post-quantum offering on the fabric.
- Every crossing encrypted under your keys
- Near line-rate encryption across clouds and regions
- Algorithm as policy, changed with a configuration push
- The Cloud Native Security Fabric is quantum safe
- Zero Trust for Cloud-Native Workloads
Aviatrix Clearance
Two controls and an autonomous defense layer, bought together or in any order, priced per workload and delivered on Aviatrix Cloud.
- Outbound Containment governs egress: only the domains a workload is cleared for
- Lateral Containment governs east-west: only the workloads it is cleared for, down to the pod and the function
- Monitor first, with the Shadow Rule Set showing what policy would have blocked
- Aviatrix Active Defense keeps clearance current as the estate changes
You are here - Zero Trust for AI Workloads
AI agent discovery and containment
See every AI agent in your cloud, including the ones no one authorized, and set what each may reach, in the network, with no SDK and no code change.
- Gateway-free discovery of AI agents from flow and DNS analysis
- Default-deny outbound for AI agents with Aviatrix-managed AI WebGroups
- Where agents share an interface, the product says so
- Runs inside Aviatrix Clearance at the agent grain
Already an Aviatrix customer? Talk to your account owner.
See what every workload can reach, free, in five minutes
The Self-Score tests five properties of your environment and reports your Blast Radius with no procurement and no budget approval. Aviatrix Clearance is free to every enterprise through May 1, so you have time to patch.