In managed hosting, the host patches the platform and the site owner patches the website. Every website runs on two stacks. Your host maintains the one underneath: the server, its operating system and kernel, the control panel, and the server software the platform provides. You, or whoever maintains your site, look after the application on top: WordPress, plugins, themes, custom code and the accounts that can log in.

Neither half covers for the other, and most unpleasant surprises in hosting happen because nobody drew that line explicitly. Providers tend to answer the question with adjectives. This is the longer answer, with the line drawn.

Updated November 14, 2026: We have added the CVE identifiers for Dirty Frag, which this article first described as “CVE pending”: CVE-2026-43284 for the xfrm ESP flaw and CVE-2026-43500 for a related RxRPC flaw. We have also corrected two details: cPanel’s EasyApache 4 fixes arrived across several releases rather than one, and the datacenter building itself is run by its operator, not by us.

Who patches what: the host maintains the Linux kernel, control panel, web server, PHP builds, database engine, caching layer, firewall and mail stack; the site owner keeps WordPress core, plugins, themes, custom code, user accounts and what gets installed.

Two stacks, one website

The half most people think of as the website is the application layer: WordPress core, plugins, themes, custom code, the user accounts with access to it, and the decision to install something in the first place. It has a dashboard. It shows update badges. It tells you when it wants attention.

Underneath it sits the platform: the Linux kernel, the control panel, the web server, the PHP builds, the database engine, the caching layer and the firewall. None of it appears in your dashboard. There is no badge, no notification, and usually no sign that anything happened at all.

Where the line falls

This is how the split works on Blue Arctic hosting. The exact line moves a little between products, so the notes column says where.

LayerWho maintains itNotes
Server hardware and networkBlue ArcticAll products. The servers and network equipment are ours; the datacenter building in Tampa is run by its operator.
Operating system and kernel patchesBlue ArcticAll products, including Cloud Servers, where root access is also yours.
Control panel and PHP updatesBlue ArcticManaged and tested by us. On a Cloud Server, software you install yourself over SSH is yours.
Malware scanning, web application firewall, SSL, edge DDoS filteringBlue ArcticSwitched on for every product before a site goes live.
BackupsBlue Arctic, where the product includes themDaily on Shared Web Hosting and Managed WordPress (kept 30 days) and on VPS (5 recovery points). Virtual Dedicated Servers include none unless you add backup storage. Keeping your own copy of anything business critical is still yours.
WordPress core, plugins and themesYouOr a maintenance plan. On Managed WordPress Hosting, Patchstack virtual patching blocks known plugin and theme flaws at the platform level, which buys time. It does not update the plugin.
Passwords, user accounts and who has accessYouIncluding turning on two-factor authentication wherever your application offers it.
What gets installed in the first placeYouThe platform can contain a bad plugin. It cannot decide not to install it.

The same split is published on our security page as what we do and what stays yours.

Product by product

The layers above are the same everywhere; how much of the server is yours is what changes. On Shared Web Hosting and Managed WordPress Hosting you never touch the server. On a VPS or a Virtual Dedicated Server you hold root, so anything you install yourself joins your half of the line.

ProductBlue Arctic maintainsStays with you
Shared Web HostingThe server, operating system, web server, PHP, databases, cPanel, malware scanning, the firewall, email, DNS and SSL. Daily backups, kept 30 days, and restores.Your website: the application, its plugins and themes, your content, and who can log in.
Managed WordPress HostingThe servers, caching, the firewall, malware scanning, Patchstack protection, and daily backups kept 30 days.Your themes, plugins and content, and when updates run. Website Care can take that over.
VPS HostingHardware, network and the virtualization layer, plus Core Managed: operating system security patches and kernel updates, Imunify360, CloudLinux, and cPanel/WHM or Plesk installation, updates and maintenance. Daily backups, 5 recovery points.Root access and everything you install over SSH, including your applications. Server Administration adds engineers who watch the applications too.
Virtual Dedicated ServersHardware, network and the virtualization layer, plus the same Core Managed coverage as a VPS. Backups are not included unless you add backup storage when you order.Root access and everything inside the server, and your own backups unless you add backup storage. Server Administration is available here too.

Retention, restores and what each product includes are set out in Section 8 of our Terms of Service and on each product page.

What a few weeks of platform maintenance looks like

Spring 2026 is a reasonable illustration, because several unrelated things landed close together.

cPanel shipped a run of EasyApache 4 releases with security and maintenance fixes across a long list of components: Apache, PHP, Tomcat, Redis, Valkey, nginx njs, ModSecurity and LSAPI. In the EasyApache 4 change log they appear as versions 25.57 to 25.61. Between them, those updates touched most of the software that actually serves a page.

Separately, three Linux kernel privilege escalations arrived within about two weeks:

  • Copy Fail (CVE-2026-31431), in the algif_aead module. CloudLinux’s advisory, published April 30, 2026, says all Linux kernels since 2017 are affected.
  • Dirty Frag, in the kernel’s xfrm (IPsec) code, disclosed on May 7, 2026 according to TuxCare, the KernelCare vendor. It was listed as CVE pending at the time and is now tracked as CVE-2026-43284, with a related RxRPC flaw tracked as CVE-2026-43500.
  • Fragnesia (CVE-2026-46300), which CloudLinux reported on May 13, 2026 as affecting CloudLinux 7h, 8, 9 and 10, with a working public proof of concept.

None of that was visible from inside a website.

Local does not mean someone in the building

Kernel privilege escalation flaws get described as local vulnerabilities, and that word does a lot of damage. It sounds like a problem for somebody with physical access to the machine, which sounds like somebody else’s problem.

On shared hosting, local means any account already on that server. That is the whole point of the vulnerability class. An attacker who gets a foothold through an ordinary web application weakness, possibly an outdated plugin on a different site on the same machine, starts as an unprivileged local user. A kernel privilege escalation is the step that turns that foothold into root, full control of the server.

That is why a flaw three layers below your application, in software you have never configured and cannot see, is your problem too, even though it is not your job to fix. It is also why account isolation matters on shared hosting. On our Shared Web Hosting, CloudLinux gives each account its own CPU, memory and process limits, and CageFS keeps each account’s files out of view of the others. Isolation narrows what a foothold can reach. It does not replace patching the kernel.

Why neither half covers for the other

The two stacks fail independently.

A carefully patched platform does nothing for a site running a plugin with a published vulnerability and no update in two years. Automated tools look for sites running known vulnerable software, and they have no idea what your site is or how small the business behind it is.

A disciplined plugin routine does nothing about a kernel bug. You cannot patch a kernel from inside WordPress, and no amount of application hygiene changes what the machine underneath is running.

That is the practical content of the word managed. Not that somebody is watching your site, but that somebody owns the half you cannot reach.

The exposure window is the real measurement

For platform vulnerabilities, the number that matters is not whether a fix exists. It is how long the gap lasts between disclosure and that fix being applied everywhere.

That gap is rarely a technical problem. What stretches it is deciding to apply the fix, finding a maintenance window, and confirming it landed on every machine. When working exploit code is already circulating, those days are the risk.

Live kernel patching is the main tool that shortens it. It applies a kernel fix to the running kernel in memory, with no reboot. On CloudLinux that is KernelCare, and KernelCare livepatches were released for all three of these issues: for Copy Fail, for Dirty Frag and for Fragnesia (in the CloudLinux advisory linked above). It was part of how we handled them.

The limits are worth stating plainly. Live patching does not cover every kernel fix, and some still require a reboot. Control panel, web server and PHP updates follow their own paths, and some need a service restart. Live patching does not abolish maintenance windows. It removes them from the kind of fix where waiting costs the most.

Four questions worth asking any host

  • Which components do you patch, named specifically, and which stay with me?
  • Do you tell me when you act on something, or do I find out by asking?
  • For kernel level issues, do you live patch, and what is the usual gap between disclosure and the fix being applied?
  • On the half that stays with me, who is actually doing it? The person who built the site four years ago is a common and uncomfortable answer.

None of these are secrets. They are just rarely on a pricing page.

Where our line sits

Our security page lists what we do on the platform side: apply operating system and kernel patches proactively, critical and high severity first; manage and test control panel and PHP updates; and monitor server health, file changes and unusual log activity around the clock. The PHP versions we offer, 8.3, 8.4 (the default) and 8.5, are all still within their security support period on the PHP project’s supported versions page.

For the kernel issues this spring, we wrote up what they were and what we did about them:

The application half stays with the site owner, and our SLA reflects that: where we tell you a security patch, update or configuration change is needed and it is not applied, service credits are excluded for problems that follow. That is not a dodge. It is the part of the boundary worth knowing before something needs patching, rather than after.

If you would rather not own the application half yourself, Website Care adds a person who reviews every WordPress core and plugin update before it goes live, on a site hosted with us or anywhere else.