The INTEL
In 2025, attackers got faster and defenders got slower. Both ends of the window moved the wrong way in the same year, and the reason on the defender’s side is volume.
The defender’s side, from the Verizon DBIR 2026:
- Only 26% of critical known-exploited vulnerabilities (those on CISA’s KEV list) were fully fixed in 2025, down from 38% the year before.
- The median time to fully fix one rose to 43 days, from 32.
- The reason: the median organization had 16 of these vulnerabilities to fix, up from 11. “That is almost 50% more.”
The attacker’s side:
- Microsoft: the median time from a flaw being found in the wild to a working exploit is “well below 24 hours.”
- CrowdStrike: once inside, the average attacker moves through the network in 29 minutes, down from 84 in 2022.
Why It Matters
A 43-day fix cycle against a one-day exploit was already a losing race. What changed this year is the cause. Security teams didn’t simply get worse: they were handed about 50% more critical fixes, and the fix rate fell.
Our read: this is why SIGNALS says “expose less,” not “patch faster.” A faster team still loses to rising volume. Fewer exposed systems means fewer critical fixes in the first place, and every fix you don’t need is one the attacker can’t race you to.
What To Do — One Key Action
Ask your security leader for two numbers from last year: how many critical known-exploited flaws the company had to fix, and what share were fully fixed. If the workload grew faster than the fixes, the answer is fewer exposed systems, not just a faster team.
MITRE ATT&CK
- T1190 — Exploit Public-Facing Application: the flaws that sit unfixed on exposed systems
- MITRE mitigation listed in SIGNALS: M1016 Vulnerability Scanning, paired with M1042 Disable or Remove Feature or Program
MITRE ATLAS (attacks on AI systems, v5.6)
- AML.T0049 — Exploit Public-Facing Application: internet-facing AI applications run the same fix race, with the same exploit clock
- AML.M0023 — AI Bill of Materials: a full listing of what each AI system is built from, which ATLAS says can “enable rapid response to reported vulnerabilities.” You can’t fix what you can’t find
- AML.M0016 — Vulnerability Scanning: extends scanning to AI model files and their software dependencies, which “should be scanned for known vulnerabilities”
Learn More
- FIR SIGNALS 2026 — Signal 01, with every figure checked at the source
- Verizon 2026 Data Breach Investigations Report — Primary source
- FIR Risk INTEL-60 — The Door You Opened on Purpose — The other half of exposure
Powered by FIR Risk Platform — AI-driven threat intelligence for enterprise risk leaders.