Gigascale KEV Analysis: Human Security Fails?
Apr 10, 2026 // 17:57 - Tristan Wall


By: Saeed Abbasi, Threat Research Unit’s Senior Manager at Qualys

With exploits appearing before patches and AI-driven threats intensifying, minor improvements aren’t enough. We need a fundamentally new approach to security.

Key Takeaways for Leaders

Analyzing CISA’s list of Known Exploited Vulnerabilities over the last 4 years reveals that even with security teams resolving 6.5 times more issues, the percentage of critical vulnerabilities still unpatched after a week actually increased from 56% to 63%. More staff alone won’t fix this.

Of the 52 actively exploited vulnerabilities studied, 88% were patched slower than they were exploited; half were actively used before a patch was available.

The bottleneck isn’t speed, but the operational framework itself.

Overall exposure, not just the number of vulnerabilities, is the real measure of risk that security teams must now track. While dashboards prioritize quick patching, attacks exploit vulnerabilities that linger. AI isn’t just another attack surface; the most dangerous time is when AI-powered attackers are up against human defenders.

Therefore, security teams must implement automated, self-regulating risk management systems.

New findings from the Qualys Threat Research Unit, based on analysis of over one billion remediation records for CISA KEV vulnerabilities across 10,000 organizations over four years, confirm what many in the industry have suspected but never fully demonstrated: the current operational model for enterprise security is flawed.

The number of vulnerabilities has risen 6.5 times since 2022. According to Google M-Trends 2026, the time between vulnerability disclosure and exploitation has shrunk to negative seven days, meaning attackers are actively exploiting vulnerabilities before patches are released. The proportion of critical vulnerabilities remaining open after a week has increased from 56% to 63%.

This isn’t due to a lack of effort. Organizations are resolving 400 million more vulnerabilities each year than before. Teams are working harder, yet the impact where it matters is minimal. Our researchers call this the “human ceiling” — a fundamental limitation that more staff or better processes cannot overcome. The problem isn’t effort, it’s the framework itself.

Of 52 significant vulnerabilities that were actively exploited, 88 percent were patched after they were already being used in attacks. For example, Spring4Shell was exploited two days before it was publicly announced, yet it took enterprises an average of 266 days to patch it.

Similarly, the vulnerability in Cisco IOS XE was being exploited a month before disclosure; the average patching time was 263 days.

The attackers had a head start measured in days, whereas the defenders took months to respond. It’s not a lack of information, but a failure in execution.

The report highlights a “Manual Tax” — the compounding effect where vulnerabilities on less-visible assets, which require manual handling, extend exposure from weeks to months. For Spring4Shell, the average remediation time was 5.4 times longer than the median.

The median gives a simplified view, but the average reveals the reality. Infrastructure systems are in a worse situation: for Cisco IOS XE, the median patching time was still 232 days — compared to endpoints, where the median is consistently under 14 days. When even the best-case scenario takes eight months, the Manual Tax isn’t just a factor, it *is* the baseline.

Relying on average figures is no longer effective for making decisions. Instead, focusing on Risk Mass — the number of vulnerable assets multiplied by the duration of exposure — provides a clearer picture of cumulative exposure than simply counting vulnerabilities. A related metric, Average Window of Exposure (AWE), measures the total time from when a vulnerability is exploited to when it’s patched across the environment.

For example, Follina was exploited 30 days before it was disclosed, with an average patching time of 55 days after disclosure.

However, the AWE extended to 85 days. While the period before disclosure accounted for 36 percent of that 85 days, the long tail of patching accounted for an additional 44 percent. Together, the time before disclosure and the long patching tail represent 80 percent of the total exposure time. The patching sprint that gets measured accounts for less than 20 percent.

Meanwhile, of the 48,172 vulnerabilities reported in 2025, only 357 could be exploited remotely and were actively being used in attacks. Organizations are wasting resources patching vulnerabilities that pose a theoretical risk while ignoring real, exploitable weaknesses.

Cybersecurity has always followed technology trends — Windows security followed Windows, cloud security followed cloud. Leading experts now believe that AI changes this pattern. It’s not just a new area to defend; it’s a fundamental change in the adversary itself.

Automated tools can already identify vulnerabilities, create exploits, and launch attacks faster than any human team can respond. The remediation data shows that human teams can’t keep up today. AI-driven attacks will widen this gap in the future.

The period of transition — when AI-powered attackers are facing human-speed defenders — is the most critical time for the industry, exacerbated by underlying vulnerabilities: attack surfaces that are too large for teams to manage, identities that grow faster than policy implementation, and patching workflows that still rely on manual processes.

The traditional scan-and-report strategy was designed for fewer vulnerabilities and longer exploit timelines. It should be replaced by an end-to-end Risk Operations Center: built-in intelligence providing machine-readable decision guidance, active validation confirming whether a vulnerability is actually exploitable in a specific environment, and automated actions shortening response times to match the speed of the threat.

The aim isn’t to eliminate human input but to improve it, allowing professionals to move from tactical execution to managing the policies that guide their automated systems.

The organizations that are succeeding in closing the exploit gap are not doing so with larger teams. They are succeeding because they have eliminated human delays from the critical workflow.

The scan-and-report strategy — find, score, assign, manually route — was designed for fewer vulnerabilities and slower attack timelines.

It should be replaced by an end-to-end Risk Operations Center: built-in intelligence providing machine-readable decision guidance, active validation confirming whether a vulnerability is actually exploitable in a specific environment, and automated actions shortening response times to match the speed of the threat.

The aim isn’t to get rid of human involvement but to enhance it — allowing professionals to shift from tactical execution to controlling the policies that govern automated systems. The organizations already winning are not winning with larger teams. They are winning because they have eliminated human delays from the critical workflow.

Exploit timelines are not going to get longer and vulnerability volume will not decrease. The reactive method has reached a point of diminishing returns.

The only question left is whether organizations will update their strategies to keep up — before the advantage shifts too far in favor of automated attacks.

Sponsored and written by Qualys.

#analysis  #fails?  #gigascale  #human  #kev  #news  #security   —   News