On 10 June 2026 CISA issued Binding Operational Directive 26-04, "Prioritizing Security Updates Based on Risk". It revokes and replaces two directives that had shaped vulnerability management far beyond the US federal government: BOD 19-02, which set patch deadlines by CVSS severity, and BOD 22-01, which created the Known Exploited Vulnerabilities (KEV) catalog. The directive binds only US federal civilian agencies, but its logic is the most useful public answer to a question every European IT team asks each week: of the hundreds of findings in front of me, which ones do I fix first? This article explains the three tools behind that answer (KEV, EPSS and SSVC), what BOD 26-04 changed, and how to turn it into a policy that fits a European mid-market company. It extends our guide to the vulnerability management lifecycle, which covers the full process from discovery to remediation.
Key Takeaways
- CVSS measures how severe a flaw is, not how urgent it is for you. CISA's own guidance now says severity labels "do not prescribe specific actions".
- KEV answers "is it being exploited?", EPSS answers "how likely is exploitation in the next 30 days?", and SSVC turns those signals plus your context into a decision.
- BOD 26-04 ranks urgency on four factors: internet exposure, KEV status, whether exploitation can be automated, and whether the attacker gains partial or total control.
- The most urgent case gets three days and a forensic check: fixing is not enough, you also look for signs that the flaw was already used.
- Exposure decides. No scoring system works without an accurate view of what is reachable from the internet and whether the vulnerability is really there.
Why CVSS Alone Fails
The Common Vulnerability Scoring System gives every CVE a number from 0 to 10 describing the intrinsic severity of the flaw. It was never meant to be a to-do list, yet for years it has been used as one: fix every "critical" in 15 days, every "high" in 30. The result is a backlog that never shrinks. The volume alone makes the point: according to ACN's Operational Summary for August 2026, as reported by ICT Security Magazine, the CVE Program published 12,719 new records that month, and 13 of them showed active exploitation.
CISA's implementation guidance for BOD 26-04 is blunt about it: severity labels such as "critical" or "high" do not prescribe specific actions, and with BOD 19-02 revoked, federal agencies are no longer required to use CVSS for prioritisation. CVSS still has a place. It describes the flaw well. What it cannot know is whether your copy of the software is exposed, whether anyone is exploiting it, and what an attacker would reach through it.
KEV: Known Exploitation
The KEV catalog lists vulnerabilities for which CISA has reliable evidence of exploitation in the wild. An entry needs a CVE ID, reliable evidence of active exploitation and a clear remediation or mitigation. A public proof-of-concept is not required. Each entry carries the date it was added, a due date for federal agencies, and a field that says whether use in ransomware campaigns is "Known" or "Unknown".
By catalog version 2026.11.10, KEV held 1,779 entries, 295 of them added in 2026. That is a short list compared with the CVE firehose, which is exactly why it is useful: if a vulnerability in your environment is in KEV, someone is already using it.
KEV has two limits worth knowing. It only lists flaws with a remediation available, so a zero-day can be exploited for weeks before it appears: the Check Point VPN flaw we covered in our article on the Qilin VPN case was exploited from 7 May 2026 and entered KEV on 8 June. And its absence proves nothing: "not in KEV" does not mean "not exploited".
EPSS: the Probability of Exploitation
The Exploit Prediction Scoring System, maintained by the EPSS Special Interest Group at FIRST, "estimates the probability that a published CVE will be exploited in the wild in the next 30 days". Scores run from 0 to 1, come with a percentile, and are published daily for every CVE.
EPSS fills the gap KEV leaves: it gives a signal for the thousands of CVEs that are not, or not yet, known to be exploited. It is a statement about the vulnerability across the whole internet, though, not about your network. A high EPSS score on a service you do not expose matters less than a moderate score on your VPN portal. Use it to rank, not to decide alone.
SSVC: From Signals to a Decision
Stakeholder-Specific Vulnerability Categorization is a decision tree rather than a score. CISA's version combines five decision points:
- Exploitation: none, public proof-of-concept, or active.
- Automatable: can an attacker automate every step of exploitation?
- Technical impact: partial or total control of the affected component.
- Mission prevalence: how central the affected system is to what the organisation does.
- Public well-being impact: the harm to people if the system is compromised.
The output is one of four actions, as defined in CISA's SSVC page: Track (no action now, standard update timelines), Track* (standard timelines, closer monitoring), Attend (supervisory attention, remediate sooner than standard) and Act (leadership attention, remediate as soon as possible). The strength of SSVC is that it forces you to answer questions about your own environment, which is what CVSS leaves out.
What BOD 26-04 Changed
The directive replaces severity-based deadlines with timelines that depend on four variables, informed by SSVC:
| Factor | Question |
|---|---|
| Asset exposure | Is the vulnerable asset reachable by unauthenticated or untrusted parties over a public network? |
| KEV status | Is the CVE in the Known Exploited Vulnerabilities catalog? |
| Exploit automation | Can an adversary automate all the steps needed to exploit it? |
| Technical impact | Does exploitation give partial or total control of the asset? |
Some details are worth borrowing even if you are not a US agency:
- Three days for the worst case. A known-exploited flaw on a publicly exposed asset, automatable, with total impact, gets a three-day deadline, and the agency must also perform forensic triage to check whether the asset was already compromised.
- Timelines are dynamic. Taking an asset off the internet can relax its deadline; a new KEV entry shortens it. The clock starts at the earlier of the KEV addition or the moment the agency finds the vulnerability on its asset.
- Exposure must be known. Agencies have to tag every externally reachable asset by environment, exposure and type. When exposure data is missing, CISA's fallback treats public IP addresses and network devices as exposed.
- Timelines are maximums. Agencies may patch faster. Agencies have 180 days from issuance to apply the new timelines in full, which, according to law firm Wiley, means by 7 December 2026.
The guidance also describes recommended forensic triage steps, counted from the KEV addition: scoping in the first 2 hours, evidence preservation and critical patching between 2 and 24 hours, containment between 6 and 24 hours, triage analysis between 24 and 48 hours and an escalation decision between 48 and 72 hours. Note what is absent: the directive does not mention EPSS.
The effect is visible in the catalog itself. In our count of KEV version 2026.11.10, 162 entries were added between 10 June and 10 November 2026. Of these, 126 carry a three-day due date and 36 a fourteen-day due date, and 103 are flagged for forensic triage.
Why It Matters in Europe
BOD 26-04 has no legal force in the EU. NIS2 does, and Article 21(2)(e) lists "vulnerability handling and disclosure" among the measures essential and important entities must take, without saying how to prioritise. In Italy, ACN's baseline measures for NIS entities follow the same logic. A documented, risk-based method borrowed from a published government framework is easy to defend in front of an auditor or a board, far easier than "we fix the criticals when we can". For the regulatory side, see our complete NIS2 guide.
A Worked Example: Two "Critical" Findings
Imagine a scan returns two findings, both labelled critical. The first is real; the second is a typical pattern described generically.
| Finding A | Finding B | |
|---|---|---|
| What | CVE-2026-50751, Check Point VPN authentication bypass, CVSS 9.3 | A CVSS 9.8 flaw in a library on an internal test server |
| Exposure | Internet-facing VPN gateway | Internal only, no route from the internet |
| KEV | Yes (8 June 2026), ransomware use "Known" | No |
| Automatable | Network-reachable, no user interaction | Requires a specific local configuration |
| Technical impact | An attacker gets a VPN session into the network | Total, on a test machine |
| Decision | Act: fix within days, and check logs for prior compromise | Track: fix in the next scheduled cycle |
CVSS ranks B above A. Every risk-based method ranks A far above B, and the difference is not a matter of opinion: one is being used against companies like yours, on a system anyone can reach.
Exposure Decides: Applying It in Practice
Every model above assumes two things you may not have: an accurate list of what is exposed to the internet, and confidence that the vulnerability is really present. A scanner that reads a version banner and lists every CVE for that version produces findings, not facts. Prioritising unverified findings with excellent math still sends your team to the wrong servers.
This is how RECON External approaches it. It discovers what your domains expose, matches detected products to CVEs weighed by CISA KEV, Exploit-DB, Metasploit and EPSS, and then tests each finding against the target, marking it Proven, To verify or Closed with the evidence of every test. Each finding is scored on severity (40%), exploitation signals such as KEV and EPSS (35%), exposure from the internet (15%) and how sensitive the technology is (10%), then weighted by its verdict: proven counts in full, to verify at 0.8, disproved and excluded not at all. KEV-listed CVEs are decided on the evidence like every other finding. RECON shows that a vulnerability exists; it does not exploit it.
For cloud accounts and SaaS tenants, RECON Cloud puts findings in order with a CISA SSVC decision that combines reachability, data sensitivity and exploitation (KEV and EPSS). A known-exploited finding gets a risk floor of 90 out of 100, and one on a reachable admin or crown-jewel resource always counts as critical.
A Policy You Can Adopt
The table below is an example starting point, not CISA's timetable. Adjust it to your resources and risk appetite, write it down and have it approved.
| Condition | Example target | Extra step |
|---|---|---|
| In KEV and internet-facing | Days, not weeks | Check for compromise from the first known exploitation date |
| In KEV, internal only | Short cycle | Review who and what can reach the asset |
| Not in KEV, high EPSS, internet-facing | Short cycle | Re-check KEV and EPSS weekly |
| Not in KEV, low EPSS, internal | Standard patch cycle | Track |
| No fix available, actively exploited | Mitigate now | Reduce exposure, increase monitoring |
Checklist
- Build and keep current an inventory of internet-facing assets, including the ones IT did not deploy.
- Tag each asset with exposure (public or internal), environment and owner.
- Pull KEV and EPSS into your vulnerability data automatically, every day.
- Verify findings before ranking them, so the queue holds real problems.
- Write a short, risk-based remediation policy and get it approved by management.
- For every KEV finding on an exposed asset, add a compromise check to the fix.
- Measure time to remediate per category, not just the number of open findings.
If you want to see which of your exposed findings are proven and how they rank on KEV and EPSS, start with a RECON External scan: the first three domain scans are free. For cloud and Microsoft 365 posture, RECON Cloud applies the same SSVC logic through read-only connectors.