In April 2026, Mozilla shipped Firefox 150 with fixes for 271 vulnerabilities that an Artificial Intelligence (AI) model had found in a single evaluation. An earlier scan using a different model produced 22 fixes. Models like this are now being used to analyze the software every enterprise runs on.

For most security teams, the bottleneck is remediation (fixing what you find quickly without breaking the systems the business runs on). AI is forcing an operating model change as it finds things at a pace that traditional security operations simply can’t keep up with. That’s where VulnOps comes in.

Why the scan, ticket, and patch cycle keeps falling behind

Most vulnerability programs still operate on a calendar. Scanners turn findings into tickets, sort them by severity score, and then let them sit until the next maintenance window. In Verizon’s 2026 Data Breach Investigations Report, exploitation of a vulnerability was the most common method of initial access (31% of all initial access, up from 20% the year before). Only 26% of the vulnerabilities in the Cybersecurity and Infrastructure Security Agency (CISA) Known Exploited Vulnerabilities catalog were fully remediated in 2025, and the median time to close them increased from 32 to 43 days. Gartner’s average exposure window is 50 days. If you don’t have the remediation capacity to deal with the volume of vulnerabilities you’re going to find, more scanning just adds to the queue.

What is VulnOps?

VulnOps is the continuous operations model for vulnerability management. The term was coined by Heather Adkins, Gadi Evron, and Bruce Schneier in an October 2025 essay, and the Cloud Security Alliance (CSA) has published a very detailed model based on Development and Operations (DevOps), the practice that brought software and operations teams together. Findings from new code, running systems, and vendor advisories go through one governed pipeline where each one is validated, and the fix is confirmed to have worked. The CSA makes it very clear that this is an organizational change rather than a tool purchase.

Several vendors now use the term. Altimetrik implements VulnOps as an engineering-led model for enterprises and runs it at scale in production. It’s built on top of the scanners and fix tools that you already have in place and takes every vulnerability through the four stages of Rank it, Fix it, Prove it, and Prevent repetition. Our distinction is how we govern the remediation process via policy gates, risk-tiered approvals, staged rollouts, and verification.

How governed AI remediation keeps your people in control

In our implementation, the AI agents do the actual remediation and humans are at the helm. 

Rank it: agents create a single trusted record of every vulnerability from across your scanners, eliminate the duplicates, and then rank each one by likelihood of exploitation and business impact. 

Fix it: AI agents and engineers then go and fix the issue wherever it is (in the app code, in third-party software, in the infrastructure, in the containers, in the cloud config, etc.). Not every vulnerability is going to require the same response, so we triage by blast radius and complexity and then pick one of four. 

Fix fast applies when the defect is bounded and the test is clear. Agents propose the fix, it’s reviewed by engineers, and then the test confirms it. When you have a pattern that’s repeated across systems, then Fix at Scale comes in to identify the variants and then roll out the adapted fixes in stages. If you have a safe fix that you need to do over a period of time, then contain then resolve provides the bridge with compensating controls that have an owner, a review date, and exit criteria. A structural weakness is a different beast, and that’s where modernize the platform comes in. In that scenario, you’re using AI to do the engineering on the components and the architecture that are affected.

Prior to anything actually getting to production, it’s checked against your policies. Low-risk change classes that your teams have already approved just go through monitoring. Anything with a wide blast radius is at a gate until an engineer or security owner has approved it. 

Prove it: every fix is tested and retested in the live environment, fails before the change and passes after it. A finding only closes as a verified fix that’s running in production or as a finding that’s been closed by a human.

The broader market is moving in the same direction. Gartner has named autonomous exposure remediation (their term for AI and automation that can identify, prioritize, and fix vulnerabilities) one of their top 10 strategic technology trends for 2027. The concept is that the system can automatically push a patch into production, and Gartner predicts that by 2031 organizations using this will reduce their average containment window from weeks down to under 48 hours. For the enterprise, the issue is all about how that autonomy is governed.

Our controls vary with the change. Cloud policies are tested in isolated environments prior to release, and third-party patches go through a staged rollout with rollback triggers.

How policy-as-code stops the same vulnerabilities from returning

Proving the fix closes the current finding. To prevent repetition, the fourth step uses policy-as-code guardrails and gates throughout the lifecycle so the same class of vulnerability does not reappear. The CSA paper points out that a lot of the findings are of the same type (e.g., a missing authorization check which has been copied across the codebase) and that patching each instance individually just leaves the root cause. We can turn those recurring fixes into policy written as code and enforce them as gates at each stage where that vulnerability class could enter, from design and source through build, release, and production. A base image that fails its hardening threshold goes back for a rebuild before anyone can deploy it.

The CSA tracks progress via an absorption rate, resolved findings divided by new findings over a rolling 30 days, and a rate above 1.0 indicates that the backlog is in fact shrinking. Guardrails improve that ratio from the intake side by preventing repeat classes from becoming findings.

Where C-suite leaders should start with VulnOps

Ask your security lead for two figures: the number of validated vulnerabilities that your teams closed last month and the number of new ones that came in. If the number of new ones is greater than the number of ones you’ve closed, then your backlog is in fact increasing. Look at both of those figures in the context of scanning scope, change failure rate, and recurrence so that you don’t end up with reduced coverage or unstable fixes being presented as progress. The CSA recommends you start with a fairly narrow pilot (one dev pipeline and one segment of running systems), and a few 30-day cycles will tell you whether the model works.

If you would like to explore what a VulnOps pilot might look like for your organization

Talk to our team


References

  1. “The zero-days are numbered,” The Mozilla Blog, April 21, 2026
  2. “2026 Data Breach Investigations Report,” Verizon Business, 2026
  3. “Gartner’s top tech trends for 2027 expand—and constrain—AI capabilities,” Network World, October 20, 2026
  4. “Autonomous AI hacking and the future of cybersecurity,” CSO Online, October 8, 2025
  5. “The VulnOps Operating Model,” Cloud Security Alliance, 2026
  6. “Gartner Identifies the Top Strategic Technology Trends for 2027,” Gartner, October 20, 2026

About the Author

Rizwan Patel
Rizwan Patel
Global Head of Cloud, Information Security, and Emerging Technologies, Altimetrik


Rizwan leads the cloud, AI, and secure architecture practice at Altimetrik. He’s a strong advocate for the use of agentic AI to improve API visibility and cyber resilience and works with clients to help them proactively detect and self-heal risks. He’s a regular speaker and writer on AI and cybersecurity and firmly believes that accountability, explainability, and trust are the foundations of any modern digital system. His background is in cloud transformation, deep learning, and governance, and he’s very much at the cutting edge of helping digital enterprises build trusted innovation in a complex threat environment.

Contact Us

We'd love to hear from you.
Contact Us