This sheet is past its review date. Deadlines or advice may have changed since 5 Nov 2026.
Patching: how fast is fast enough?
If a vulnerability is being exploited and the system is reachable from the internet, contain it within 24 hours and patch within 72. Schedule everything else by exploitation and exposure, not by CVSS score alone.
Patch deadlines by risk
Our recommended deadlines, counted from when a fix or official workaround is released. Use the first tier that matches.
Exploitation starts here: an estimated 7 days before a fix exists, on average1
-
Exploited and exposed
Known exploited, on a system reachable from the internet. Edge devices first.
Contain in 24 h Patch within 72 hours -
Exploited, internal
Known exploited, reachable only from inside your network.
Patch within 7 days -
Exposed and severe
Critical or high severity, or a high EPSS score, on an internet-facing system. No known exploitation.
Patch within 14 days -
Internal and severe
Critical or high severity on internal systems.
Patch within 30 days -
Everything else
Medium and low severity, no known exploitation.
Patch within 90 days
If you only do five things
Know what faces the internet.
Keep a live inventory of internet-facing systems. A 72-hour deadline is useless on a server nobody knows about.
Patch edge devices first.
Security and networking products were hit by 21 of the 43 zero-days aimed at enterprise technology in 20257. If the flaw was exploited, also check them for compromise.
Automate the routine.
Enforce automatic updates for operating systems, browsers and apps centrally, so people can focus on the few urgent cases.
Pre-approve the emergency path.
Decide in advance who may patch or isolate a system outside the change window, so the 72 hours aren't spent waiting for a meeting.
Triage a new vulnerability
Is it being exploited?
Check CISA KEV6, the vendor advisory and your national CERT. Vendor wording like "limited, targeted exploitation" counts as yes.
Can it be reached from the internet?
Check your external inventory. VPNs, firewalls, gateways, mail and file-transfer systems count as exposed by default.
What breaks if it's compromised?
Identity systems, edge devices, backups and anything holding customer data move up one tier.
How likely is exploitation soon?
For everything else, use EPSS8 to break ties and recheck weekly. Then record the tier, the deadline and one named owner.
What changed in 2026
- CISA replaced BOD 22-01 with BOD 26-04 in June 2026. From December 2026, federal deadlines depend on exposure, exploitation, automatability and impact, and the highest-risk cases get 3 days3.
- NVD stopped enriching every CVE. Since April 2026 it prioritizes KEV entries, software the US government uses and critical software9. Take severity from the vendor or CNA, and combine it with KEV and EPSS.
- EPSS moved to version 5 on 15 June 20268. Recheck any score thresholds you tuned on older versions.
- EU reporting duties for manufacturers started on 11 September 2026 under the Cyber Resilience Act10.
Can't make the deadline?
- Cut the exposure. Put the system behind a VPN or an IP allowlist, or switch off the vulnerable feature.
- Apply the vendor's workaround and write down when it stops being acceptable.
- Watch the attack path. Add WAF or IPS rules and alerts for the known exploit.
- Give the exception an end date. An accepted risk has a named owner and expires.
What you can skip
- Sorting by CVSS alone. It rates severity, not whether anyone is exploiting the flaw. Most published vulnerabilities are never exploited.
- Chasing zero findings. Medium and low findings on internal systems belong in the regular cycle. Track deadlines met per tier, not the total open count.
- Emergency patching every headline. If it isn't exploited and isn't exposed, it goes in the regular cycle.
- Long scan reports. A one-page list of overdue first- and second-tier items does more than a 300-page export.
Mistakes that cost the most
- Patching without restarting. Many fixes aren't live until the service or device restarts.
- Stopping at the patch. Patching won't remove an attacker who is already in. On exploited edge devices, check the vendor's indicators of compromise, rotate credentials the device handled and review logs from before the patch.
- Forgetting what scanners miss. Appliance firmware, container base images and libraries bundled inside other products.
- Reporting averages. A good average can hide the one VPN that stayed unpatched for 6 weeks. Report the slowest cases.
- No owner for vendor-run systems. If a supplier operates it, the patch deadline belongs in the contract.
What this looks like at your size
Small team, no security staff
- Enforce automatic updates from device management.
- Where you can, use cloud services instead of self-hosted VPNs and gateways, so the provider patches the edge.
- Subscribe to advisories for the few products that face the internet.
- Name one person who owns patching, plus a backup.
Growing, with IT or security staff
- Keep an external attack-surface inventory that updates itself.
- Enrich scanner findings with KEV and EPSS.
- Write down your tiers and deadlines and get them signed off.
- Get an emergency change path pre-approved.
Enterprise or regulated
- Report deadlines met per tier to leadership every month.
- Run risk acceptance with owners and expiry dates.
- Put patch deadlines and advisory duties into supplier contracts.
- Map your tiers to NIS2, DORA or PCI DSS, and keep a compromise-assessment playbook for edge devices.
A separate clock: EU reporting
These run alongside your patch deadline and can start the moment you learn a flaw is being exploited.
Cyber Resilience Act
Manufacturers, for actively exploited vulnerabilities in their own products. Applies since 11 September 2026; reports go through ENISA's Single Reporting Platform1011.
- 24 hoursEarly warning
- 72 hoursVulnerability notification
- 14 daysFinal report, counted from when a fix or mitigation is available
NIS2
Essential and important entities, for significant incidents12. NIS2 sets no patch deadline; its implementing rules for digital service providers ask only for patches "within a reasonable time"13. A signed-off tier table shows what reasonable means for you.
- 24 hoursEarly warning
- 72 hoursIncident notification
- 1 monthFinal report, counted from the notification
Measure the right things
- Deadlines met per tier. Aim for 100% on the first tier.
- Oldest open known-exploited vulnerability. Track its age in days.
- Time from advisory to decision. Slow triage eats the deadline before anyone patches.
- Coverage. The share of systems reporting to your patch and scanning tools. Systems you can't see never show up as overdue.