PHP 8.2 end of life on December 31, 2026: what actually happens to your website—and the checklist to run before it

The last security release for PHP 8.2 ships December 31, 2026. Two remote-code-execution flaws from October show exactly what stops arriving in January—here is the runbook.

The short answer

PHP 8.2 receives its last security fix on December 31, 2026. After that date the branch is dead—no more security releases, ever, for any site still running it. If your site is on PHP 8.2, that deadline is about six weeks out. The rest of this article is the plan.

This is not a corner-case deadline, either. PHP runs 68.2% of all websites whose server-side programming language is known, and PHP 8.2 alone still accounts for 25.9% of PHP 8 usage—behind only 8.3. Roughly one modern PHP site in four is sitting on this exact date.

What end of life actually means—and why it is not “deprecated”

PHP’s support lifecycle has three states, and php.net defines all three on one page. Active support: reported bugs and security issues get fixed, and regular point releases ship. Security fixes only: critical security issues get patched, and releases happen, in php.net’s own words, “on an as-needed basis.” End of life: the release is “no longer supported,” and users of it are told to upgrade “as soon as possible,” because they “may be exposed to unpatched security vulnerabilities.”

PHP 8.2 has been in that middle state since the start of 2025—active support ended December 31, 2024. On January 1, 2027 it drops into the third state, and nothing in the project’s plan ever picks it back up.

“Deprecated” is the word that gets tangled into this, so untangle it once. Deprecation is a code-level notice: PHP flags a function or a behavior as deprecated to tell developers it will be removed in some future release, and the site keeps working meanwhile. End of life is a support state about the interpreter itself. Nothing deprecates your PHP 8.2 on December 31—it just stops receiving defense.

The current PHP 8 branches, with their dates from php.net’s supported-versions table:

PHP 8.2—released December 8, 2022. Active support ended December 31, 2024. Security fixes end December 31, 2026.

PHP 8.3—released November 23, 2023. Active support ended December 31, 2025. Security fixes until December 31, 2027.

PHP 8.4—released November 21, 2024. Active support ends December 31, 2026. Security fixes until December 31, 2028.

PHP 8.5—released November 20, 2025. Active support until December 31, 2027. Security fixes until December 31, 2029.

PHP 8.6—general availability scheduled for November 19, 2026 (release candidate 7 was out for testing this week).

Put that table next to a calendar and December 31, 2026 does double duty: it is the day 8.2 loses security fixes entirely, and the day 8.4 drops from active support into the security-only lane. Six weeks earlier, on November 19, 2026, PHP 8.6.0 is scheduled to go GA. Three version events inside six weeks is why every host and agency you follow is suddenly writing about PHP.

The October proof: two remote-code-execution flaws, fixed while 8.2 was still alive

To feel what stops arriving in January, look at what arrived in October. PHP 8.2.35 shipped as a security release, and php.net’s announcement says it in two sentences: “The PHP development team announces the immediate availability of PHP 8.2.35. This is a security release.” Two advisories were fixed, and both are the kind you move dates for.

CVE-2026-109618 (GHSA-w8hw-w9fx-j79w): rated Critical, CVSS 9.2—remote code execution through out-of-memory corruption. When a request hits memory_limit, PHP bails out of the request with longjmp(), and in some contexts request memory “is temporarily reachable before it is fully initialized.” The PHP Foundation’s October 21 post on the release is blunt: “On Apache with mod_php, an unauthenticated request can trigger this,” and “Remote code execution was also confirmed on PHP-FPM through application-level code paths.”

CVE-2026-109619 (GHSA-6wmf-4rv5-p3rp): remote code execution via stale global state—effectively a use-after-free on a shared worker process.

Both fixes shipped in the same window on every live branch: PHP 8.5.12, 8.4.27, 8.3.36, and 8.2.35 carry the same two advisory IDs in the changelog. That parity is what a supported version buys you—a flaw appears, and whichever branch you run gets the patch.

The advisory record for CVE-2026-109618 lists every 8.2 release below 8.2.35—8.2.0 through 8.2.34—as affected, with 8.2.35 as the patched version. If that range includes your site today, the fix exists and is one upgrade away.

Now run October forward into March 2027, with your site still on 8.2. An equivalent flaw turns up. The fix ships in 8.3, 8.4, 8.5, and 8.6—and nothing ships for 8.2, because there is no 8.2.36 and never will be. That is the whole answer to “what happens if my website runs PHP 8.2 after end of life”: nothing changes on the night, everything changes after it.

The two-minute self-check: what version is your site actually running?

Before any planning, find out what you actually run—not what your 2024 invoice says, what the server answers today. Three ways, fastest first.

Open your hosting panel. Nearly every panel—cPanel, Plesk, CloudPanel, and the custom dashboards managed WordPress hosts build—shows each site’s PHP version as a per-site setting. In WordPress, Tools → Site Health → Info lists it too.

Drop a one-line probe file into your webroot, load it in the browser, read the version banner at the top, then delete the file.

Ask whoever runs your server, precisely and in writing: “Which exact PHP version does this site run in production, and when did it last change?”

<?php
phpinfo();

Call the file something like php-check.php and delete it as soon as you have read it. phpinfo() prints your whole server configuration—paths, modules, environment—and that is exactly the page you do not want sitting in a webroot for six weeks.

With SSH access, the tempting check is one command:

php -v

Treat that with suspicion on shared hosting: the command line often runs a different PHP binary than the web server. The probe file is the truth for your site; the terminal is the truth for your shell. When they disagree, believe the file.

Why hosts force the upgrade silently—and how that breaks sites

On shared hosting, one interpreter serves thousands of tenants, so an unpatched PHP branch on that box is the host’s liability too. Hosts solve it the same way every cycle: as a version nears end of life, they pick a date and move every site on it up a branch. Some announce it. Many do it quietly. Either way, your site changes its PHP branch overnight, with no staging pass and nobody watching your forms at 9 a.m.

What breaks is rarely WordPress core—WordPress.org recommends “PHP version 8.3 or greater,” with a stated floor of 7.4, so core runs comfortably across every live branch. What breaks is the long tail of plugins and themes written against old behavior: a constructor style from years back, a changed function signature, an assumption that warned on 8.2 and fatals on 8.3. Your site does not break because the new PHP is bad; it breaks because a plugin vendor stopped testing long before your host flipped the switch.

Notice, too, that “recommended” is not “newest,” and neither means “safest.” WordPress’s stated target, 8.3, has itself been security-fixes-only since the start of 2026. Aim a business move at 8.4 or 8.5 instead: both still get active fixes today, with security fixes into 2028 and 2029.

The alternative to a silent bump is a planned one—staging, a test pass, a date, and a way back. That sequence is the core of managed WordPress hosting: the update arrives on a schedule someone chose for your site deliberately.

PHP 8.6 and the ABI break, in plain terms

PHP 8.6.0 is scheduled for November 19, 2026—three days from today, and six weeks before 8.2’s deadline. As of this week it is at its seventh release candidate, one more than the plan called for, and the reason is worth understanding.

The October security fix changed 8.6’s internals. The internals report on the 8.6 release puts it flatly: the fix “broke the ABI in 8.6 only,” so extensions “must be rebuilt.” ABI means application binary interface—the binary contract compiled extensions use to talk to the interpreter. Not function names: struct layouts, offsets, calling conventions. The October fix moved those, which turns every compiled extension built against an earlier 8.6 build—a caching handler, a database driver, an image library—into a binary making promises the interpreter no longer keeps.

For most readers this is not your job—your host or your panel builds those extensions. What it changes is the shape of the jump: a major PHP upgrade is not a click but a window in which every binary extension gets rebuilt, smoke-tested, and watched. Schedule it like one, and do not chase 8.6 on launch day anyway. Move 8.2 sites to 8.4 or 8.5, which have years of runway, and let your host ship 8.6 first.

The runbook we run on our own stack

Here is the order I move a site across PHP branches. Our stack is CloudPanel on Hetzner with Ubuntu 24.04 LTS underneath, and CloudPanel supports PHP 7.1–8.5 there—so the worked example targets 8.5, not day-one 8.6.

Snapshot the VM, and separately take a file-plus-database backup you can restore without the snapshot. Two ways back, not one.

Clone the site to staging and switch staging to PHP 8.5 first. The PHP version is a per-site setting in CloudPanel, so staging can run the target branch while production still answers on 8.2.

Update everything on staging—WordPress core, plugins, theme—before judging anything about PHP. Outdated plugins produce noise that looks like PHP breakage.

Exercise the money pages: submit every form end to end, run a test order through checkout if you sell anything, fire the scheduler manually (wp-cron on WordPress) and watch it finish, and send one test mail from the site.

Read the error logs—staging first, then production for the first day or two after the switch.

Deploy out of hours: flip production to 8.5, keep the old 8.2 settings written down for a fast fallback, and expect your page cache to serve some stale content until its TTL runs out.

Honest about scope: on a normal brochure or store site this is an afternoon. On a custom plugin stack it is a multi-day project with real discovery. What it is never: a dashboard checkbox you hit during a coffee break.

When the updater itself needs updating

October produced a second, quieter lesson, and it belongs in every maintenance plan. CVE-2026-109518 is a flaw in APT itself—the package manager that installs Ubuntu’s security updates. An attacker positioned on the path between your server and its package mirror “can trick the client into installing an attacker-controlled unverified package and achieve remote code execution with root privileges.” Ubuntu rates it High and shipped the fix for 24.04 LTS as apt 2.8.4ubuntu0.1.

Sit with that for a second: the tool that applies your patches was itself the vulnerable package. It got fixed the way everything else does—by an update—but only if something installed it. This is why “unattended-upgrades is on” is not a maintenance strategy. Automation applies updates; it does not verify they landed, and it cannot notice when the updater itself is the thing that needed patching.

On our stack the check is one read-only command—Ubuntu 24.04 LTS (noble) carries the fix as apt 2.8.4ubuntu0.1:

apt policy apt

Read the Installed line: 2.8.4ubuntu0.1 or later means the box that updates everything is itself current. Anything older, and the December checklist just got a new first item.

The seven questions to email your host this week

Whether you manage the server or someone else does, these questions surface a silent-bump policy before it finds your checkout. Each comes with what a good answer sounds like—and what a red flag sounds like.

“Which exact PHP version is my site running in production right now?” Good: an exact release number, plus the date it last changed. Red flag: “the latest”—that is not a version.

“What happens to my site when PHP 8.2 reaches end of life?” Good: a dated migration plan with notice. Red flag: “we always keep you current,” with no date anywhere in the reply.

“Can I set the PHP version per site, and switch back if something breaks?” Good: yes, per site, with a self-service or same-day path back. Red flag: the version is tied to the server as a whole.

“Can I stage the site on the new PHP version before production changes?” Good: a staging copy or dev URL running the target version. Red flag: “just test on production during low traffic.”

“Which PHP versions do you support, and until when?” Good: a published list with end dates. Red flag: “current versions,” with no list.

“Where are my site’s error logs, and does anyone read them after a PHP change?” Good: self-service logs plus a named check after the flip. Red flag: logs exist, but nobody opens them unless a customer complains.

“Are OS and server package updates automated, and how do you verify they applied?” Good: automation plus a verification step you can see. Red flag: “updates just happen.” The APT flaw above is exactly why this question exists.

A host that answers in dates and version numbers is one you can plan December with. A host that answers in reassurance is how a PHP change lands on your site without warning—and this check costs one email.

Your December 31 checklist

Six weeks from today, these should simply be true:

You know the exact production PHP version—checked this week, not recalled from an old invoice.

If it was 8.2: a tested move to 8.4 or 8.5 is deployed, the money pages are exercised, and the old settings are written down as a fallback.

WordPress core, plugins, and theme are current, because you updated before judging the new branch.

A snapshot and a restorable backup exist, dated after everything above.

Somebody reads the error logs for the first days after the switch and knows how to roll back if they fill up.

A maintenance plan exists that names the next PHP move before it becomes an emergency.

That last item is the real deliverable. PHP’s cadence is public and boring—two years of active support, two of security fixes, then the branch is gone, with no grace period and no exceptions for your plugin vendor. A plan that treats December 31, 2026 as scheduled maintenance, not a fire drill, is what keeps the next October from becoming your outage.

And a next date is already on the calendar. On February 10, 2027, every Let’s Encrypt subscriber moves to 64-day certificate lifetimes by default; the authority expects “the last 90-day certificate to expire on May 11, 2027,” and will shrink the authorization reuse period from 30 days to 10. “We will not revoke valid certificates as a part of this process”—nothing breaks on the day—but if your renewals are automated and your client supports ACME Renewal Info (ARI), “you should be all set,” as long as it renews at roughly two-thirds of the certificate’s lifetime rather than the old rhythm. We walk through that switchover in a separate post on certificate lifetimes.

If question one above already made you nervous, run the self-check tonight—two minutes tells you which side of December 31 you are on. And if you would rather hand the whole runbook to a team that runs it every month, that is the work we do. Tell us about the site—we will start with the same question every plan starts with: what is it actually running?