November 19, 2026
The new world order: every bug is critical, so none of them are
This week went something like this. Tuesday: Java4Shell. Wednesday: Nginx2Root. Both rated 9.2, both with a catchy name, and both with a…

By Charlie Kelly
4 min read
This week went something like this. Tuesday: Java4Shell. Wednesday: Nginx2Root. Both rated 9.2, both with a catchy name, and both with a wave of launch-day blogs telling you the sky was falling. For a lot of security teams that meant dropping everything to work out whether they or their customers were actually exposed.
Then PHP pre-announced a critical security release for all supported versions, due Thursday. (No name yet, give it a day). Funnily enough, PHP's announcement is the one that got it right: it said up front that the critical issues need non-default configuration or uncommon setups. One sentence. That's all it takes.
A few well-known voices in the industry said the quiet part out loud this week, and I agree with them. The bugs are real. The way they're being launched is hurting the people they're supposed to help.
The problem is the launch, not the bug
Somewhere along the way, vulnerability disclosure turned into a product launch. A bug now shows up with a brand name, a "2shell" suffix, sometimes its own domain, a launch-day blog and a sales team who'd love to hop on a call. What it often doesn't show up with is a clear list of what has to be true for it to work.
I spend my days watching people register domains they probably shouldn't. So forgive me for noticing when a vulnerability gets its own domain faster than it gets a list of preconditions.
Defenders pay for all of this. When every bug is critical, teams can't tell "patch this week" from "drop everything," so they drop everything. Every time. Do that often enough and people stop listening, and the next bug that really is everyone's problem gets the same tired shrug as the last five that weren't.
The timing doesn't help either. If you're in the UK, you know the routine by now. The vuln goes public at 8pm, while everyone's having their tea. You log on at 9am and there are already demands to explain your exposure and have a comm out. Then it's a frantic scramble to match half the internet's hot takes against a sprawling tech stack. And repeat.
Don't overcorrect
This isn't a "these bugs are fake" post. Some of them absolutely get exploited. Last month's PHP2Shell, another leaked runtime RCE scored 9.2, has since been added to CISA's Known Exploited Vulnerabilities catalog.
So the ask isn't "calm down". It's "give us context". Tell us what has to be true for the exploit to work, and let defenders decide how fast to move.
The two examples from this week
Nginx2Root. Five nginx bugs, disclosed on November 18. The headline one lives in HTTP/2, and a server is only in play if an attacker can negotiate HTTP/2 with nginx itself. HTTP/2 might not even be turned on in your config. And if nginx sits behind a CDN that terminates the connection, exploitation is far less likely. As of November 19 there was no reported exploitation and it wasn't in CISA KEV. It's fixed in 1.30.7 and 1.31.9. Oh, and the bottom of nginx2root.com? A sales form.
Java4Shell. It affects OpenJDK 17.0.6 and newer on x86–64 CPUs with AVX-512 VBMI support. So not every Java box you own, then. At disclosure, Calif said the bug had leaked and that technical details would follow Oracle's patch. Meanwhile, neither Oracle nor OpenJDK had published an advisory naming the CVE, the CVE record was only reserved, and there was no known exploitation or public PoC.
I'll let you draw your own conclusions on that one.
The score won't save you
CVSS 4.0 already has an Attack Requirements metric, built for exactly this: the conditions that have to be in place on the target for an attack to work. And yet Nginx2Root, Java4Shell and PHP2Shell all came out at 9.2. Needs a specific CPU? Needs a specific config? Doesn't matter. Critical is critical.
Dave's "oh shit meter" is a nice idea. Until something like it exists and people actually use it, the context has to come from whoever discloses the bug, in plain words, up front.
Nobody is coming to triage for you
You'd hope NIST would step in here. It can't, and it has said so. Since April 15, 2026, the National Vulnerability Database only enriches the highest-risk CVEs, and NIST has stopped routinely adding its own severity score, relying on the one from the CVE Numbering Authority instead. Around 29,000 backlogged CVEs were reclassified as "Not Scheduled".
Vendor programs aren't in much better shape. Microsoft has written that when reports arrive faster than MSRC can resolve them, you get a growing queue (Microsoft). It patched a record 200-plus vulnerabilities in June as AI-assisted discovery ramped up.
That's not a dig at NIST or MSRC. They're great teams. They're just short on people and drowning in reports, like everyone else in this pipeline. It just means the context has to come from whoever discloses the bug, because nobody downstream has time to add it.
What Researchers & Vendors should do
- Don't overhype; lead with the preconditions. Config, modules, CPU features, network position. In the first paragraph, not paragraph twelve.
- Show it working somewhere realistic. Not in a lab built to make it work.
- Give a realistic likelihood, and say what you don't know yet.
- Drop the "2shell" suffix and the vanity domain. The CVE number has done the job for decades.
- Coordinate the fix before the fanfare. Responsible disclosure first, launch party second.
- Hold the sales calls until people have at least finished reading the advisory.
What to actually do
- Patch everything internet-facing on a fast, boring, routine cycle, whatever the headline says. Do that and most of these weeks become a non-event.
- Put an edge in front of your origins, and lock the origins down so they only accept traffic from it. Cloudflare's CTO, Dane Knecht, said Cloudflare's edge was fixed globally before Nginx2Root was disclosed, so proxied traffic locked down to Cloudflare needed nothing from customers. If you use it for DNS only, you still need to patch your origin, but everyone should be using Cloudflare anyway.
- Know your exposure. Is HTTP/2 actually on? Which modules do you run? Is this box even internet facing?
- Save the war room for bugs that are exploited in the wild, land in KEV, or are reachable in your environment. Everything else goes in the normal cycle.
Welcome to the new world order
AI-assisted discovery isn't slowing down, so there will be more of these weeks, not fewer. Honestly, that's fine. More bugs found means more bugs fixed. But if every one of them launches like the end of the world, defenders will tune them all out, and the one that really matters will get the same eye-roll as the rest.
Find the bug. Ship the fix. Tell us exactly what it takes to exploit. Skip the logo.