Philadelphia Live News

collapse
Home / Daily News Analysis / The internet runs on PHP. AI just found a big problem. Should you be worried?

The internet runs on PHP. AI just found a big problem. Should you be worried?

Oct 26, 2026  Twila Rosenbaum  66 views
The internet runs on PHP. AI just found a big problem. Should you be worried?

Key facts at a glance

  • Two critical PHP vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, received emergency patches after researchers showed they could be exploited through a single script.
  • Both flaws carry a severity rating of 9.2 out of 10 and can allow remote code execution on vulnerable servers.
  • The PHP Group issued an alert on October 20, 2026, and released fixes in PHP 8.5.12, 8.4.27, 8.3.36, and 8.2.35.
  • Attackers began probing the flaws before patches were broadly available, and a major content delivery network reported blocking millions of exploit attempts within hours of disclosure.
  • AI tools helped researchers uncover and weaponize the vulnerabilities, shortening a process that might once have taken months into roughly a week.

What is PHP and why does it matter?

PHP is a general-purpose scripting language that has been a workhorse of web development for decades. It runs on servers, allowing websites to generate dynamic content, handle forms, manage sessions, produce HTML and PDFs, and serve data formats such as JSON and XML. Many popular publishing platforms rely on it, including WordPress. Industry tracking data has estimated that nearly 70 percent of websites that use a server-side language depend on PHP, and that around 40 percent of all websites run on WordPress. Those figures explain why a serious PHP flaw is not a niche server-room problem. It can affect a huge share of the public web, from small blogs and volunteer group sites to government portals and large commercial services.

Why the PHP vulnerabilities became urgent

On October 20, 2026, the PHP Group published an alert for upcoming security updates designed to patch two very high-risk vulnerabilities. The flaws were later identified as CVE-2026-109618 and CVE-2026-109619. Both allow an attacker to run remote code, and both carry a severity rating of 9.2 out of 10. According to the PHP Group, the flaws affect all PHP installations serving web requests. The existence of the vulnerabilities was known before the alert, but they were widely assumed to be too difficult to exploit in practice. Researchers at the cybersecurity firm Wiz challenged that assumption. They found a way to exploit the flaws through a single script, which was named PHP2Shell. The research was developed with partners including OpenAI, and it changed the risk calculation almost overnight.

The advisory set off a rush across the web. Server administrators began preparing for patched versions of PHP. When details of the two vulnerabilities were released on October 21, websites began going offline temporarily as upgrades were completed. Government websites in the Netherlands, France, and Germany were among those affected. The disruption was a sign of how quickly a server-side flaw can become an operational emergency. Patching such a widely used language is not as simple as updating a single app. It involves coordinating hosting providers, content management systems, custom applications, and server configurations. Each layer can introduce delay, and each delay creates a window for attackers.

Why this is a big deal

Two factors make the PHP flaws especially serious: the number of affected sites and the speed at which modern vulnerabilities can be exploited. Hypothetically, if the flaws had been left unpatched, more than half the internet could have been at risk of hacking. That is not a theoretical concern for server administrators alone. A compromised website can spread malware, steal data from the server, delete data, or be used as a launch point for further attacks. Even a site that appears small can become part of a larger criminal infrastructure if it is left vulnerable.

The speed factor is equally important. In the pre-AI era, vulnerabilities could take days, weeks, or even months before a working exploit was seen. Attackers also spread their techniques more slowly, and defenders had more time to patch. In the current AI era, exploits can be deployed rapidly, sometimes in mere minutes. AI did not just accelerate the threat after disclosure. It helped researchers uncover the vulnerabilities in the first place. Researchers at Wiz used frontier AI models, including Google Gemini 4 Argon and OpenAI GPT-5.6 Sol, while uncovering the flaws and developing working exploits. According to Wiz, AI helped turn vulnerabilities that had lurked in PHP for more than a decade into something a single researcher could exploit in a week. The company estimates that the same task could previously have taken a world-class security team months.

That is the double-edged sword of AI in security. On one side, AI can help defenders uncover dangerous bugs before attackers do. On the other side, it rapidly lowers the time and expertise needed to turn obscure flaws into working attacks. The PHP2Shell research illustrates both effects at once. It showed that a long-standing assumption about exploit difficulty was wrong, and it gave defenders a narrow window to patch before attacks scaled. That window did not stay open for long.

Attack attempts escalated quickly

As predicted, attacks escalated quickly. According to ongoing coverage, a major content delivery network spotted millions of attempts to exploit the vulnerabilities, which were dubbed PHP2Shell. Amazon security researchers detected signs of attackers probing the flaws as early as October 17, four days before emergency patches landed. Hackers were also reportedly tweaking their attacks to slip past defenses. By the evening of October 21, the content delivery network reported already blocking millions of attempted exploits. Its CEO said the numbers began spiking as soon as the flaw disclosures went live.

Firewalls and other defenses helped, but they were not impenetrable. Coverage does not extend to all corners of the internet. The content delivery network in question is used by about 21 percent of all websites, leaving many sites outside its protection. Attackers also continue to adapt their techniques, rapidly iterating with the help of AI. That combination of broad exposure, fast exploitation, and incomplete defensive coverage is why the PHP flaws drew such urgent attention. For website owners, the message was clear: patch now, and do not assume that a lack of visible symptoms means a site is safe.

Which PHP updates include the fixes?

The PHP Group directly issued patches within versions 8.5.12, 8.4.27, 8.3.36, and 8.2.35. If a website relies on an older version, it must receive the necessary security fixes through a vendor that helps maintain that older version. Administrators should verify the exact PHP version running on their servers, not just assume that a host has applied the update. Some consumer web hosts are quietly handling this work in the background. In one case, a volunteer group's WordPress site had already been upgraded to 8.3.36, and a support team message said it was globally applying updates and had configured its firewall to block exploit attempts. That is the ideal outcome, but it is not guaranteed for every site.

For people running PHP as part of a NAS, a Raspberry Pi, or another home server setup, the responsibility is more direct. They should put on their server administrator hat and ensure a proper update. The same applies to anyone maintaining a custom application, plugin, or theme that depends on PHP. Updating the core language is essential, but it may not be enough if an application bundles outdated components or has its own security gaps. A complete response includes checking logs, reviewing installed software, and confirming that firewalls and monitoring are active.

How ordinary users may be affected

Some websites may go offline or remain offline until they are running a secure version of PHP. But as of the afternoon in the United States, many sites were operating normally. Most ordinary users should be able to visit sites as usual. That does not mean the risk is zero. Even with a server's firewall defenses in place, a website using an unpatched version of PHP could be attacked and used to run remote code. In that scenario, the site could spread malware, steal data from the server, or delete data on the server. Users may never see a warning, which is why standard security habits still matter.

When browsing the internet, the best defense is the same standard advice. Keep browsers and antivirus software up to date. Share only necessary information with websites. If you store data on a site, ensure you have backups of that data just in case. These steps do not fix a vulnerable server, but they reduce the harm if a site you use is compromised. They also help limit the damage from phishing pages, malicious ads, and other common threats that often follow major security incidents.

How to check if a WordPress site is safe

WordPress users can check their PHP version from the administration dashboard. In the left-hand navigation bar, look for Tools, then choose Site Health. Two tabs will appear in the main part of the screen. Select Info, then choose the Server section. A drop-down screen should reveal the PHP version. If the site is not running one of the patched versions, the owner should contact the hosting provider or the vendor responsible for maintaining that version. If the site is on a managed host, the host may have already applied the update, but it is still worth verifying.

Site owners should also check for signs of compromise. Unusual log entries, unexpected file changes, unfamiliar administrator accounts, and unexplained traffic spikes are all reasons to investigate. The absence of obvious symptoms is not proof of safety. Attackers often try to remain quiet after gaining access, and a compromised site can be used for days or weeks before the damage becomes visible. For administrators, the safest immediate step is to verify the PHP version, apply the patched release, and review logs for suspicious activity dating back to October 17, when early probing was detected.


Source: PCWorld News


Share:

Your experience on this site will be improved by allowing cookies Cookie Policy