Aviatrix Clearance: Solving the CISO Catch-22

Most of the CISOs I talk to know the problem only too well: network security is one of the most critical layers in the cyber-security stack, but their workloads in the cloud are poorly protected. Adding network security in a brownfield environment risks breaking production, and if done in the way they’re familiar with, it will slow down the app teams. They are working as hard as they can to prevent security incidents, but the backlash for causing downtime may be worse than the fallout from a breach. 

This tension makes a robust, long-lasting, efficient security program extremely difficult to implement, but rewriting the entire script is too expensive and disruptive for the board to agree to. The result is frustration and an ongoing security dilemma that makes chasing vulnerabilities and sending “out-of-compliance" emails the default. Teams hold their breath as breaches become a weekly headline.

Four Points of Friction 

I’ve seen four points of friction that prevent CISOs from fully protecting their networks: 

Where to Start 

One of the things I say frequently is that I hate the word “Zero Trust.” It implies perfection, and I’ve never seen a project implemented to perfection. In fact, if perfection is the only goal, many teams will never start. Analysis paralysis is common, and the key is figuring out: which workloads should we protect first? On-prem, we called them “crown jewels,” but those sound risky. In the cloud, we should take a few additional factors into play: which workloads have sensitive data? Which segments have the biggest exposure? Which ones can access the internet?

Many enterprises have wide-open egress internet access, a gift to attackers. But in a sprawling brownfield environment, there are too many workloads to tackle at once. 

Inserting Safely 

How can you insert a solution that doesn’t break production? If you don’t have a policy enforcement point in place, just the act of deploying something feels risky. You’re either adding a hop in the network, or deploying an agent. Either way, you’re touching production. Here’s where scope, granularity, performance, and quickly reversible actions make a difference. The number one fear is that inline enforcement breaks production applications, something no CISO wants to touch. 

Safe Enforcement 

As with insertion: how can you enforce security policies without blocking legitimate traffic? Almost every mid-sized organization runs applications that they don’t understand, let alone understand well enough that they’d feel comfortable going from a wide-open access environment to something that uses a granular allow-list for segmentation and security. Tools and processes for mapping applications remain immature. I still see customers using logs and spreadsheets to attempt manual discovery. For a global corporation with thousands of connections across time zones, blocking a necessary connection could feel worse than a security incident.

Maintaining the New Reality 

Even if an enterprise managed to insert and enforce safely (or close enough that it was worth it), the day two maintenance of adding new applications, managing migrations, and getting approvals is even worse. Did the team miss something that was important for running end-of-quarter financial reports? What happens when the app team needs a new service added? Do they revert to the legacy two-week window for change management? Is the security team staffed for the increased volume in change requests? Did we introduce too much friction to achieve the broader business objectives? 

The Growing Urgency: We Can’t Kick This Can Any Farther Down the Road 

Many security leaders have struggled with this reality for years, and it’s caused inaction, but the danger is growing. In the hands of an attacker, they make patching vulnerabilities impossible. Last week, PHP announced an industry-shaking RCE vulnerability. The team even knew that it could be risky, but assumed that any exploit would be too complex for attackers to discover. AI discovered the vulnerable chains and was able to build PoC exploits in under 24 hours. After publishing fixes, teams discovered that word about the vulnerability had leaked. As much as 4 days prior to a patch being available, attempted exploits were detected running on the internet. 

Every B-tier attacker now has the power of a nation-state, and every normal organization needs to rethink how they defend against it. Hint: you can’t patch fast enough. Internally, AI agents represent a Toxic Combination for enterprises. They’re ephemeral, highly privileged, and ship much faster than security has time to review. Many enterprises haven’t figured out who owns them yet, much less how to control them. 

I was walking through RSA two years ago, reflecting that every vendor on the show floor wouldn’t exist if organizations didn’t have to connect to the internet. It’s both necessary and the thing that introduces the most risk. Traditionally, it was the connection that we spent the most time securing, but concepts like “Identity is the new Perimeter” have led to way too many enterprises taking the internet for granted, especially in the cloud. I see far too many enterprises that have wide-open egress that a compromised agent could use to exfiltrate data in minutes.

While people often think of securing ingress, egress is similarly one of the most critical pieces of your security program. JPMorgan Chase technology leaders argued recently that aggressive filtering of outbound traffic from production systems “would have substantially mitigated the impact of incidents like Log4Shell, SolarWinds, and numerous supply chain compromises.”

Even last week, we saw multiple examples. With eMudrha, effective egress filtering can immediately revoke trust enterprise wide for the compromised signing certificate, with php2shell, many of the steps after initial exploit would involve downloading a malicious piece of software – an egress flow, with LiteLLM round 2, the first call home was to a domain that would have been quickly blocked (if not pro-actively blocked) by most firewalls. All of them are attacks that affected workloads, and unfortunately, far too few workloads have effective egress coverage in the cloud. 

Aviatrix Clearance: Default-Deny Enforcement that Doesn’t Break Production 

In aviation, flights can only launch when they have the right clearance. Our newest offering, Aviatrix Clearance, provides that kind of transparent inspection and authorization for cloud workloads, without threatening to break production. Aviatrix Clearance is the security buyer’s offering on the Aviatrix Containment Platform, built on the Cloud Native Security Fabric. Read more.

Aviatrix Clearance offers both speed and safety by starting everyone with free visibility and a risk assessment, and a quick path to enforcing egress security. Next, you’ll get a list of connections that would be blocked if you turned on full enforcement, so you can make sure they don’t cause disruption. Then, you can enforce, blocking malicious or unknown traffic while still permitting the traffic your business needs to run.

We’ve designed this to answer each of the challenges I described above. Each step is progressive, reversible, and free, allowing you to see the security you can achieve without causing your company downtime. 

See

First, we’ll show you what your Blast Radius is with a free Containment Assessment. 

Phase 1 is self-scored and takes five minutes, free to anyone. It returns a provisional Blast Radius and a harvest exposure read.

Phase 2 gives you an Architecture Read. It’s read-only, agentless, starts with read-only config analysis, and then optionally measured from production flows, and returns five measured reads including egress coverage with the NAT dependency map, Blast Radius for lateral movement highlighting risky paths in their business context, even with a firewall in place it measures incremental risk created by cloud-native workloads, and finally Reachable Value at Risk in dollars. It measures and reports each surface, outbound and lateral, separately. It also makes an inventory of AI agents across your organization.

Insert

Next, insert without breaking production by using NAT gateway replacement, or transparent insertion (with AWS TGW), one toggle per VPC, route tables orchestrated, original addresses preserved, and a one-action backout. Hybrid enforcement will orchestrate native security groups and Kubernetes NetworkPolicy where a gateway is not needed.

Most customers start in monitor mode, but you can also start in enforce mode. You’ll receive a Shadow Rule Set report of everything Aviatrix Clearance would block so you can review it before enforcing anything. 

Enforce

Now you can start acting on what you know. You can promote one rule at a time, one VPC at a time. Aviatrix Distributed Cloud Firewall engine runs the function catalog: firewalling, segmentation, default-deny outbound, IPS and virtual patching, TLS inspection, and coming soon, AI guardrails. The first enforcement is the buy, whichever control it is. This part is priced by workload. 

Operate

Getting to day 1 enforcement is not enough. What I’m most excited about with this launch is Aviatrix Active Defense. Active Defense Agents keep clearance current as your estate changes, inside guardrails set by your team. You approve anything risky and can back out at any time. 

Aviatrix Clearance runs on the network you already have, including AWS Transit Gateway. It doesn’t need agents or an expensive rip-and-replace design. It’s delivered on Aviatrix Cloud and priced per workload. The same two controls run on Aviatrix Enterprise, priced per node. It’s safer, simpler, faster, and affordable, designed for security leaders who need to address their egress problem today. 

The Three Areas You Can Control Through Aviatrix Clearance

With Aviatrix Clearance, you can take control of the three areas where your network is most vulnerable: outbound traffic, lateral movement, and agentic AI. Here’s what you’ll receive: 

1. Outbound Containment – Traffic can only go where it’s cleared to go. Policies are enforced by domain instead of IP address. 

2. Lateral Containment – No central firewall: traffic can only travel between your workloads when it’s cleared to do so. These policies also work by domain instead of ephemeral IP addresses. 

3. Active Defense – Aviatrix Active Defense discovers and contains AI agents. Active Defense Agents propose, allow, route and block inside guardrails your team sets, so governance runs on its own, staying current even as your estate changes. 

What is your Blast Radius? How many agents are operating in your environment right now? Which of your workloads can reach the internet? Answer these questions and explore a solution with a free Containment Assessment. 

References 

Aviatrix Threat Research Center, “Eight Certificates That Can Impersonate Anyone: The eMudhra Key Failure and Your Real Attack Surface,” Oct. 21, 2026, https://aviatrix.ai/threat-research-center/research-insights/emudhra-tls-cert-compromise-attack-surface/

Aviatrix Threat Research Center, “php2shell: Two Critical PHP RCEs, a Pre-Patch Leak, and the PHP You Didn’t Know You Had,” Oct. 21, 2026, https://aviatrix.ai/threat-research-center/research-insights/php2shell-two-critical-php-rces-a-pre-patch-leak-and-the-php-you-didnt-know-you-had/

Bleeping Computer, "Amazon says PHP2Shell exploitation began 4 days before patches," October 22, 2026, https://www.bleepingcomputer.com/news/security/amazon-says-php2shell-exploitation-began-4-days-before-patches/

JPMorgan Chase, Fortifying the Enterprise: 10 Actions to Take Now for AI-Ready Cyber Resilience, April 2026. https://www.jpmorganchase.com/about/technology/blog/fortifying-the-enterprise-10-actions-to-take-now-for-ai-ready-cyber-resilience 

PHP, “News Archive - 2026,” https://www.php.net/archive/2026.php