Most people who run an application security programme will recognise this. The backlog grows every quarter. The team works harder every quarter. The gap between what gets found and what gets fixed keeps getting wider anyway.
This feeling now has numbers behind it. Black Duck’s 2027 AI Readiness Report analysed 65.5 million application security findings across roughly 13,100 applications at 533 organisations between June 2025 and May 2026. In June 2025, organisations were discovering 11 new findings for every one they closed. By May 2026 the ratio was 18 to 1, and it peaked at 22 to 1 in December. In April 2026 alone, 10.4 million unresolved findings were added to backlogs.
By the end of the study, only about one finding in ten had been closed.
The attacker’s clock has changed
For most of the past decade, defenders had time on their side. In 2018, the mean time between a vulnerability’s disclosure and its first exploit was 2.6 years. According to the Zero Day Clock, it is now under a day. Frontier AI models are speeding this up further. Exploits that once took specialist skill and weeks of effort can now be produced quickly and cheaply.
Remediation hasn’t moved at the same speed. At the median, findings from static analysis (SAST) closed in 28 days and open source findings from software composition analysis (SCA) in 39 days. Those are the fast ones. Findings from dynamic testing took 316 days, and findings from manual testing took 232. When exploitation runs on an hourly clock and remediation on a monthly one, every open finding is a window an attacker can use.
Easiest first, most dangerous last
The most uncomfortable finding is about order. If severity decided what got fixed first, critical vulnerabilities would close fastest. In the highest-volume testing methods, they don’t. Critical open source vulnerabilities took a median of 58 days to close, against 22 days for high-severity ones. In SAST, high-severity findings closed more slowly than medium or low.
Some of that reflects harder fixes. A critical flaw in an open source component may need a major version upgrade, while a low-severity fix is often a simple patch. But the overall pattern is clear: when findings arrive by the million, teams fix what is easiest. Nothing deliberate is setting the order.
Fixing criticals first would be an improvement, but severity alone isn’t the answer either. In the vulnerability advisories Black Duck’s researchers published in 2025, known exploits existed for 45% of high-severity vulnerabilities, 39% of medium-severity ones and 31% of low-severity ones. AI models can also chain lower-severity weaknesses into serious multistep attacks. A medium-severity flaw with a working exploit in a reachable part of a production application can be more urgent than a theoretical critical one. “Lower priority” can no longer mean “never”.
Less public intelligence to lean on
Prioritisation is also getting harder for another reason. In April 2026, NIST announced that it would prioritise enrichment for only a narrow subset of CVEs in the National Vulnerability Database. Most new disclosures will now arrive without the analysis that many programmes rely on, just as AI-driven discovery pushes disclosure volumes up. Teams that triage from NVD scores alone will need another source of vulnerability intelligence.
What the faster teams do differently
The data also shows that remediation speed tracks operational practice more closely than vulnerability volume. Organisations that ran SCA scans 10 to 29 times a month closed open source findings in an average of 30 days, against 66 days for those scanning less than monthly. That is a correlation, not a magic lever. Frequent scanning usually means testing is wired into pipelines, with findings routed straight to the developers who can fix them.
The largest organisations in the dataset also show what deliberate prioritisation looks like. Those with more than 10,000 employees closed critical and high-severity findings in a median of 12 days, against 19 days for all findings. Several smaller bands showed the reverse, taking longer on their most serious issues than on everything else.
Where to start
For practitioners, the report points to a handful of practical steps:
- Check your own data. Are your critical findings actually closing before your low-severity ones? If not, find out why.
- Prioritise by exploitability, reachability and business context, not severity scores alone. Set timeframes for lower-severity findings too.
- Track three numbers together: median time to close, percentage of findings closed and the discovery-to-closure ratio. A fast median means little if only a fraction of findings ever gets fixed.
- Close coverage gaps. Two-thirds of applications in the dataset were tested with a single method, yet manual testing, the smallest source of findings, turned up critical or high-severity issues more than half the time.
- Automate what can be automated, such as dependency upgrades and validated code fixes, and keep people for the judgement calls.
No amount of headcount will close a gap that grows by millions of findings a month. This is an operating-model problem, not a staffing one. The teams that keep up will be the ones that decide deliberately what to fix first, and then fix it before an attacker gets there.





