Emergency Fix Released for a Serious Flaw in PHP, the Code Behind Many Websites

PHP developers released emergency security updates on 21 October 2026 to fix two serious flaws. The flaws, CVE-2026-109618 and CVE-2026-109619, could let an attacker take control of a server over the internet without logging in. Help Net Security reported the emergency releases. PHP had already scheduled its critical security release for 14:00 UTC that day.

What was found

Wiz Research disclosed the flaws. It published its write-up on 20 October under the title "PHP2Shell: AI uncovers critical RCE vulnerabilities." The report described flaws in the core of PHP. One allows remote code execution, meaning someone can run their own commands on a machine they do not control. The other is a memory disclosure flaw, similar to the Heartbleed bug, which lets outsiders read data from a server's memory.

According to Wiz, the first bug was found on 19 August 2026 using Wiz Atlas, Gemini 4 Argon and GPT-5.6 Sol. The remote code execution flaw was found entirely by AI agents, which are programs that carry out tasks on their own. Wiz's research team explained this in a follow-up discussion published the same day.

Why it matters

PHP is a free, open-source programming language. It sits behind roughly 70% of websites, doing work that visitors never see. A flaw in its core therefore reaches many web hosts, content management systems (the software that runs blogs and online shops), and custom applications. Many of these are maintained by small teams with limited time to install updates.

Some organizations took their sites offline until the flaws were patched, according to TechCrunch's coverage. Reuters reported on 22 October that hacking attempts surged as companies and governments rushed to patch. It also reported that Cloudflare had seen millions of exploit attempts after the PHP disclosure. An exploit attempt is a try to use a flaw to break into a system.

Wiz has monitored these attempts since disclosure. In a follow-up analysis published on 27 October, titled "PHP2Shell: Exploitation Attempts and Remediation Trends," the company reported that none of the attempts have succeeded so far.

Google Cloud also published a response on 20 October. Wiz is owned by Google. Wiz's CTO of Exposure Management and AppSec, Tomer Schwartz, said the company worked with OpenAI, the research group Calif, and the PHP team to fix the bugs. Schwartz was interviewed by Rebecca Bellan for an episode of TechCrunch's Equity podcast on PHP2Shell and what the episode called the "coming vulnpocalypse." Wiz is also hosting a 20-minute technical threat briefing for organizations assessing their exposure.

How the fix reached users

The coordination matters most to people who run websites. Before the details became public, the flaw had to be shared privately, a fix had to be built, and the fix had to reach hosting companies. Wiz's account places the discovery in August and the public disclosure in late October. The emergency patch window that followed was short.

Applying a fix takes several steps: the software maker releases it, hosting companies install it, the applications are tested, and services are restarted. The broader context here is that these steps, not the discovery itself, are where most of the time went.

Think of it like a car recall. The manufacturer can design the fix in a day, but the fix only helps once every owner hears about it and brings the car in.

In my view, the more lasting question is what this does to the cost of managing security flaws. If AI agents can find serious flaws in a language this widely used, the hard part shifts from finding them to installing fixes across many systems that nobody has fully listed. Those who know exactly which PHP versions they run, and can roll out a patched version within hours, will likely handle this with little trouble. Those who cannot will face the same problem the next time a flaw is disclosed.

The broader context is also encouraging, with a caveat. The discovery method Wiz described uses the same kind of AI tools the industry has debated for two years. Here, it found bugs in shared infrastructure, and the community could fix them together with the language's maintainers. So far, monitoring has found no successful attacks, and the patches are in the hands of the people who run the sites. The open question is whether AI-assisted discovery will move faster than the slow, unglamorous work of patching. That will be decided by the people who run the sites, not by the disclosure itself.