CVE-2026-119618 Apache is the most urgent software flaw of the week: a critical bug in the world's most widely used web server that can let an attacker run code on a server without logging in. The Apache HTTP Server project fixed it on 11 November 2026 in version 2.4.71, and security teams are now racing to patch before attackers work out how to use it. The flaw sits in mod_http2, the module that gives Apache its HTTP/2 support, and it affects every Apache release from 2.4.55 through 2.4.70 when HTTP/2 is switched on. Red Hat rates it 9.8 out of 10 on the CVSS severity scale, close to the maximum, because the attack needs no password, no user interaction and only a network connection.

What CVE-2026-119618 Apache Is and Why It Scores 9.8

The bug behind CVE-2026-119618 Apache is a race condition, which means two parts of the server step on each other when they run at the same time. It lives in a function called assign_headers() inside mod_http2. When Apache handles HTTP/2 requests, worker threads are supposed to use their own private memory, and never touch memory owned by the connection thread. The flawed code breaks that rule for one request shape: it writes a synthetic Content-Length header into a table shared with the connection thread. That write can corrupt the heap, the server's working memory, and heap corruption can be turned into remote code execution, letting an attacker run commands with the privileges of the Apache process. Apache classifies the flaw as concurrent execution using a shared resource with improper synchronisation, known as CWE-362, and rates it critical. Red Hat's CVSS 3.1 score of 9.8 reflects the worst case: network attack vector, low complexity, no privileges required and no user interaction. The score is high because the conditions for attack are so easy to meet.

Which Versions of Apache Are Affected by CVE-2026-119618

Every Apache HTTP Server release from 2.4.55 through 2.4.70 is vulnerable, but only when HTTP/2 is enabled through the Protocols directive, using h2 for HTTP/2 over TLS or h2c for cleartext. Servers that only speak HTTP/1.1 are not affected, and the httpd default is HTTP/1.1, so many installations are safe by default. The standalone mod_h2 module published on GitHub, versions 2.0.8 through 2.0.42, is also affected, and that project has been archived with no fixed release, so anyone running it must move to the mod_http2 bundled with httpd 2.4.71. Red Hat notes that RHEL 8 and earlier are not affected because their mod_http2 predates version 2.0.8, while RHEL 9 and RHEL 10 are affected but rated Important rather than Critical, because each HTTP/2 stream there is allocated from a separate allocator, which makes exploitation much harder. Debian has fixed the flaw only in unstable so far, in apache2 2.4.71-1, while bookworm, trixie and forky remain vulnerable. The fix landed in a single commit on 11 November, the same day 2.4.71 shipped, and it changes the write target of one line so the header goes into the worker's own request pool instead of the shared table.

How the Race Condition in mod_http2 Works

To understand the bug, it helps to know how HTTP/2 differs from the older HTTP/1.1. HTTP/2 multiplexes many streams over a single connection, so one connection can carry dozens of requests at once, each handled by a different worker thread. Apache's design gives each worker its own memory allocator precisely so threads do not interfere with each other. The assign_headers() function broke that design in one narrow case, writing into a table that belongs to the connection thread. When two threads raced, the write could land in memory that another thread was using, corrupting the heap. A public reproduction lab for CVE-2026-119618 Apache that appeared on 12 November drives exactly this trigger in a container and shows the corruption on version 2.4.70 under AddressSanitizer and ThreadSanitizer, while 2.4.71 stays clean under the same traffic. The lab proves memory corruption, not a finished weapon: no public exploit code is linked from the Apache advisory or the CVE record. Red Hat goes further than other sources, saying public research has demonstrated a reliable exploit against a default-configured server with mod_http2 enabled, though it does not name that research. The safe reading is that exploitation is plausible and probably close, even if no working attack has been published.

Who Found the Flaw and How Big Is the Exposure

CVE-2026-119618 Apache was found twice, independently. Anthropic's OSS Scanner reported it to Apache on 26 October 2026, under the identifier ANT-2026-MAPKT6MZ, and Anthony Weems and Simon Scannell of Google Cloud Vulnerability Research reported it separately on 3 November. Apache credited both in its advisory, and the fix shipped sixteen days after the first report. The exposure is wide. Wiz, the cloud security company, reported on 11 November that 64% of the cloud environments it observes contain resources running a vulnerable Apache version, and 51% have publicly exposed instances. Wiz itself notes those figures count the presence of vulnerable versions, not whether HTTP/2 is actually enabled, so not every counted server is exploitable. Even so, the numbers describe a very large patching surface, and a compromised web server in a cloud setting could let an attacker move sideways to adjacent services or cloud metadata endpoints. As of 12 November there were no reports of exploitation in the wild, and the flaw was not in CISA's Known Exploited Vulnerabilities catalog. That is the position now, but it can change fast once exploit code appears.

How to Fix CVE-2026-119618 Apache Now

The fix for CVE-2026-119618 Apache is straightforward. Upgrade to Apache HTTP Server 2.4.71, or install the fixed package your distribution publishes, then restart Apache so the patched code is running. Red Hat shipped backports on 11 November: RHEL 9 takes mod_http2-2.0.26-7.el9_9.1 and RHEL 10 takes mod_http2-2.0.29-5.el10_3.1. Debian users on unstable can move to apache2 2.4.71-1, while users on bookworm, trixie and forky should watch for their distribution's fix and use the workaround below in the meantime. Container images built before 11 November 2026 will contain a vulnerable version, so rebuild them rather than assuming a fresh pull is safe. If you cannot patch immediately, disable HTTP/2: remove h2 and h2c from the Protocols directive and set Protocols http/1.1 in each configuration file and virtual host block, then restart. That closes the vulnerable path entirely, at the cost of the speed benefits HTTP/2 brings. Do not assume a server updated for the earlier HTTP/2 flaw, CVE-2026-23918 in 2.4.67, is protected: versions 2.4.67 to 2.4.70 are all affected by this new flaw. Check your version with httpd -v, confirm whether mod_http2 is loaded, and treat any server on 2.4.55 through 2.4.70 with HTTP/2 on as a patch-today job.

What Happens Next

The next few weeks will decide how bad CVE-2026-119618 Apache becomes. Security researchers will keep probing the flaw, and the history of critical web-server bugs says public exploit code usually follows a disclosure like this within days. Apache's 2.4.71 release also fixed two lesser flaws, a crash in mod_md and a one-byte stack overwrite in mod_vhost_alias, so upgrading covers more than one problem. Downstream vendors will keep shipping fixed packages, and the Debian stable releases should follow. The deeper question is how many servers never get patched at all: Apache has powered a large share of the web for three decades, and plenty of its installations run untouched for years. For those servers, the only defence is the workaround of switching HTTP/2 off. The flaw's discovery also says something about how security research is changing: one of the two finders was Anthropic's automated OSS Scanner, a machine hunting for bugs in open-source code, working alongside human researchers at Google. That combination found a flaw that had sat in Apache's code since 2.4.55, and it found it before attackers did. The race now is between patching and exploitation, and the patch is already available. Our guide to protecting yourself online explains the basics that sit behind fixes like this, and the emergency patch for the php2shell flaw in October showed the same pattern: a critical bug, a fast fix, and a short window to apply it. India's CERT-In warning about AI agents the same month was another reminder that the flaws attackers chase are the ones in the software everyone already runs.