A month ago, the SANS Internet Storm Center raised Infocon to Yellow for the first time in a few years, in response to the php2shell vulnerability. Since then, the headlines have kept coming: a major Microsoft 365 outage, a critical Java vulnerability, and a critical NGINX vulnerability disclosed on the same day. The Storm Center did not go back to Yellow. That restraint is the lesson for executives.
What the data shows
Three serious events landed in a few days, and the pressure to declare an emergency was real. The Storm Center's Java-specific honeypot was taking many attacks, yet none of the observed activity exploited the new Java flaw. Substantial attack volume and confirmed exploitation of a specific vulnerability are different things. Leaders who collapse the two into one will spend scarce capacity on the wrong problem.
The baseline matters too. Infocon Green means the Internet is "broken as usual," and teams should expect a normal 9-to-5 day. Some organizations use the Infocon level to justify staffing and overtime, so a threshold decision becomes a budget decision.
Key findings
1. An alert level that is always high stops meaning anything. The Storm Center keeps its status conservative because teams can't stay on guard indefinitely. If every critical vulnerability triggers emergency mode, you lose the ability to recognize the one that truly warrants it.
2. Recurring critical vulnerabilities are the new normal. The goal isn't to beat every exploit to every system. It's to reduce exposure, prioritize systems attackers can reach, and patch through a repeatable process. You won't patch a flaw like the NGINX issue across a large organization before an exploit appears, so a plan built on beating the exploit will fail by design.
3. Patching needs to be an engineered capability. Automate it as far as possible, starting with less critical systems at the bottom of your priority list. Automated patching carries two known risks: a failed patch can crash a system, and a patch applied incorrectly can leave a vulnerable system counted as patched. Automation can still make sense when you verify the result. A system that might be patched is better than one known to be unpatched, but don't mistake the status for proof. And taking a system down yourself beats having ransomware do it.
4. Exposure reduction limits what an unpatched flaw can do. Too many systems sit on the internet that don't need to. Restricting which data and resources each application can reach shrinks the blast radius. Supply chain dependencies belong in the same conversation. Ask whether a third-party tool such as Optimizely delivers enough value to justify the risk of it being compromised.
5. Capacity, not awareness, is the constraint. Security professionals across industries and countries commonly lack enough time, tools, and sometimes skills. Adding urgency to work the team can't currently do doesn't create capacity. Tooling, resourcing, and a sustainable pace do.
What this means for your team
Your readiness isn't measured by how fast people drop everything. It's measured by whether the program keeps responding without burning out. A team that sprints at every headline will eventually miss the one that matters, because its people are exhausted and its priorities are scrambled.
This also changes what you ask your security leader. "Are we on alert?" is a weak question. Better ones: what conditions trigger an elevated posture, what actions follow at each level, and who approves the cost? If an alert level can authorize overtime, it needs written criteria, or it becomes a lever anyone can pull.
The same discipline applies to AI. Used well, it handles repeatable work and organizational gaps, such as inventorying and tracking scripts the team already writes. Used carelessly, it adds risk. Direct it at specific jobs your people do poorly or lack time for, and set rules for what data it may touch.
Action items by priority
Do first
- Write escalation criteria. Define what separates ordinary internet risk from an exceptional event. For each level, specify the actions, the staffing implications, and the person who can authorize them. Include evidence requirements, such as confirmed exploitation in the wild versus a disclosure alone.
- Protect the plan. Tell your team in advance that a headline alone does not displace committed work. Define who may reprioritize and on what evidence.
Do next
- Stand up automated patching for lower-criticality systems. Identify the systems at the bottom of your patch priority list. Pair automation with verification so a "patched" status is confirmed rather than assumed, and accept in writing the risk that an occasional patch will fail. Testing can follow deployment on these systems rather than precede it.
- Inventory internet exposure. Find systems that are reachable from the internet and don't need to be, and remove that access. Review application permissions and cut access to data and resources each application doesn't require.
Do over the next planning cycle
- Review third-party dependencies. For each embedded tool, weigh its business value against the consequence of its compromise.
- Assess sustainable capacity. Ask your security leader what the team lacks in time, tooling, and skills, and fund the gaps. Treat the answer as a standing budget question, not a post-incident reaction.
- Direct automation and AI at defined gaps. Pick repeatable tasks, assign an owner, and set data-handling limits before rollout.





