PHP2Shell Shows How AI Is Tightening the Open-Source Security Patch Race

PHP2Shell made a familiar security problem harder to ignore: vulnerability discovery can move faster than the software supply chain’s ability to deliver fixes. When two flaws in PHP itself surfaced, organizations from Germany to the Netherlands took public services offline while waiting for patches. Germany pulled several federal sites, France temporarily shut prefecture websites, and Belgium’s foreign ministry disabled its sites. Similar precautions followed elsewhere. The outages showed how quickly uncertainty about a core runtime can affect services that rely on it.
The vulnerabilities, CVE-2026-109618 and CVE-2026-109619, became known as PHP2Shell. PHP maintainers released fixes for every community-supported version on the morning of October 21. Yet an estimated 40% of PHP users run versions the project no longer supports, according to a Zend report. That leaves a large installed base dependent on vendors, internal teams, or upgrade plans to close the gap.
As someone who has contributed to open source for more than 20 years, leads the Laminas Project, works as a Principal Product Manager for Perforce Zend and OpenLogic, and serves on the Board of the PHP Foundation, I see PHP2Shell as a warning about the pace of the next phase of open-source security. The response was unusually coordinated. The conditions that made a fast response necessary are likely to recur.
Why PHP2Shell reached beyond individual applications
PHP2Shell involves remote code execution risks tied to out-of-memory conditions and PHP’s error handling. At least one of the flaws affects every PHP version going back to 5.4. Their location matters: these defects sit in the runtime, below application code, rather than in a particular website, framework, or content management system.
That makes ordinary application defenses an incomplete answer. Wiz Research reported that an attacker could execute code remotely or expose sensitive data held in process memory, including passwords, authentication tokens, API keys, and certificates. The risk does not depend on which PHP application is running.
Datadog Security Labs independently reproduced the vulnerabilities and found that application-level authentication could not stop the attack: PHP handled the malicious request before application code ran. In its test of CVE-2026-109618, the request “never reaches the target application’s code.” Strong application security practices still matter, but they cannot correct a flaw in the runtime underneath them. The practical response was to patch PHP.
A coordinated release, under pressure
After the full impact became clear in August, the PHP Ecosystem Security Team, the PHP Foundation, and PHP release managers worked across supported versions. They also coordinated with long-term support vendors, including Zend, so fixes for older releases could arrive alongside community updates. The cross-ecosystem effort gave organizations using end-of-life versions a route to security fixes for these specific flaws.
That cooperation was tested when the CVE identifiers leaked on October 20, forcing releases forward by a day. By 14:00 UTC on October 21, all community-supported PHP versions had received updates. LTS providers also issued patches for versions no longer covered by the PHP project. This was an important demonstration of what coordinated maintenance can deliver, especially for software supporting government, health care, finance, and education.
Still, a coordinated release is a response to a single event. Keeping older installations secure over time requires sustained maintenance capacity and a clear route for users to obtain fixes. PHP’s estimated unsupported share makes that challenge tangible: even a well-prepared patch effort cannot eliminate the work of identifying deployments, assessing dependencies, testing updates, and rolling them out.
AI changes the research timeline
PHP2Shell also reflects a shift in who can carry out difficult security research and how quickly. Under OpenAI’s Patch the Planet initiative, researcher Calif identified a critical PHP flaw using GPT-5.6 Sol. PHP maintainers led the fix and release, while researchers and security providers helped with testing, technical support, and disclosure coordination. The model contributed to finding the issue; the ecosystem still had to validate it and produce usable patches.
OpenAI says PHP2Shell is among several significant vulnerabilities it has found and is addressing, and expects similar findings to increase sharply in coming months. That forecast points to a change in workload for open-source maintainers. More findings can help surface defects that might otherwise stay hidden. They also create more reports to triage, reproduce, coordinate, and fix.
The tension is not new. Security teams have long raced to disclose vulnerabilities, ship patches, and prevent exploitation. AI may compress the discovery stage, leaving maintainers and users less time to respond. The impact on maintainers could extend well beyond PHP if the volume of credible findings rises faster than project response capacity.
Unsupported software turns patching into a planning problem
For teams on PHP 8.1 or earlier, the immediate choices depend on their environment: upgrade to a fixed, supported release or obtain security fixes from a distribution or long-term support provider. A runtime upgrade can involve framework and library changes, extension compatibility, application updates, and production testing. That work is often necessary, but it can’t always happen in the hours after disclosure.
Long-term support can give teams time to plan and test a move while continuing to receive selected security fixes. It should be treated as a bridge, not a reason to leave systems on unsupported software indefinitely. The same challenge reaches beyond PHP. Spring vulnerability reports have increased in recent months, while Node.js and Python disclosed critical vulnerabilities on the day PHP2Shell emerged. Linus Torvalds has also said AI-related reports are making the Linux security list difficult to manage.
For organizations, the operational basics now carry greater weight: know which runtimes and versions are deployed, understand who supplies fixes, and maintain a response process that can assess and apply them. Teams should also identify applications that cannot be upgraded quickly and decide how they will receive security support in the meantime. Better vulnerability management starts with an accurate picture of what is running.
Response capacity must catch up
PHP2Shell should not be read as evidence that PHP suddenly became less secure. The joint work of maintainers, the PHP Foundation, researchers, and LTS providers showed that the ecosystem could coordinate a broad release under pressure. The more consequential lesson is that organizations need a durable plan for software whose official support has ended, and open-source projects need the capacity to keep pace with a faster stream of discoveries.
AI is making some security investigations more practical and speeding up parts of the research process. That has value when findings reach maintainers early and come with the support needed to verify and fix them. But faster discovery alone does not make systems safer. The next test for open source will be whether patch delivery, upgrade planning, and vulnerability response can move at a pace closer to the research that now feeds them.
The post PHP2Shell was only the beginning: AI is accelerating the patch race appeared first on The New Stack.