Your engineers are patching code they never wrote
Monthly CVE disclosures doubled this year as AI joined the bug hunt, and most of the noise sits in layers your team didn't write. Here's what it costs in hours, and how to stop owning the base layer.
Run a scanner on almost any container image and you get a long list. Critical. High. Medium. Pages of it.
Now look at where those findings live. Most of them aren't in your code. They're in the operating system packages and runtime under it. Your team didn't write that code. But your team gets the ticket.
The list only gets longer
2025 set a record, with 48,175 CVEs published. 2026 left that pace behind. Google's threat intelligence team found that monthly disclosures doubled this year, from 5,045 in January to 10,740 in August.
AI is a big part of the story. AI tools now hunt for flaws on their own, and they're good at it. In June, FIRST, the group that forecasts CVE volume each year, said AI-assisted discovery made an extreme year likely. The flaws AI finds tend to be the serious kind, too. Google found that half of AI-discovered vulnerabilities lead to remote code execution, against 26% across all CVEs.
Every one of them can land in a package that sits in one of your images.
And most teams already carry plenty. Datadog's 2026 report found that 87% of organizations run at least one known exploitable vulnerability in their deployed services. That's not a sign of bad teams. It's what happens when you start from a public base and the packages keep moving.
They do keep moving. Datadog found that the median dependency is now 278 days behind its latest major version, up from 215 days a year earlier. Nobody chose to fall behind. Updating the layers under your app is just never the top ticket.
Most of the noise isn't the real risk
Here's the part that stings. A lot of what lands in the queue never needed to be there.
Datadog took every vulnerability with a critical score and added context. Is it running in production? Is it exposed to the internet? Is anyone likely to exploit it? After that, just 18% were still critical. Less than one in five, the same as the year before.
Google saw the same thing at a much bigger scale. Of everything disclosed in 2026, only about 1 in 431 was ever seen being exploited.
So the real work is finding the few that matter. But to find them, someone still has to sort through all the rest.
The bill comes in hours
That sorting isn't free. It comes out of your engineers' week.
Checkmarx surveyed 2,350 security leaders and developers this year and found that developers spend nearly half their time, 49%, on security.
Not all of that is base image triage. But a lot of it is the same loop. A scan fails. Someone reads the finding. Someone figures out which layer it lives in. Someone bumps a package, rebuilds, retests and waits for the next scan. Then it happens again next week.
Try your own numbers. Even small inputs add up fast.
Those hours are the ones you hired people to spend on your product. On a platform team, they're the hours that should go to the clusters, the pipelines and the next thing the business asked for.
Stop owning the base layer
You can't make the CVE feed slower. You can choose who owns the layers under your app.
That's what Ghost is for. Ghost is one catalog your whole software factory builds from:
- Base images, hardened and kept current, so your team isn't patching them by hand.
- Language runtimes for the services you ship, kept current too.
- The top 3,000 packages from PyPI, npm and Maven Central.
- Signed, with an SBOM on every image and package, so security reviews start with answers.
Your engineers stop working through findings in packages they never touched. The findings left in the queue are in your code. That's work worth doing.
It fits what you already run. Point your Kubernetes deployments and Helm values at Ghost images, or mirror the catalog into Amazon ECR. Your CI/CD keeps working the way it does today. For most services the change is one line:
# before
FROM node:22
# after
FROM ghost/node:22
And the price doesn't grow with your image count. Ghost is one flat annual price for your whole organization, with no per-image or per-seat fees.
Count the hours first
Before you change anything, find out what the noise costs you now. We'll run a Ghost Container Economics Assessment. We map your images and where they run. We size the CVE work your teams carry today. Then we put it next to a Ghost proposal and prove it on a few of your own images.
Ask for an assessment and pick "A Container Economics Assessment" on the form.