PHP2Shell: AI helped turn decade-old PHP bugs into an exploit that threatens 70% of websites. Patch now
Two critical flaws in PHP itself, the language behind WordPress, Drupal and Joomla, let attackers take over servers without logging in. AI models made them practical to exploit, and attacks began before the fix was public.
By ScoviAI Team
Key points
- Patch released by PHP on October 21, 2026; tracked as CVE-2026-109618 and CVE-2026-109619 (CVSS up to 9.2)
- Every PHP version of the last decade, back to 5.4, is affected by at least one flaw
- Wiz says frontier AI models made the complex exploit chain practical 'in hours'
- Attackers are already exploiting it; Cloudflare saw millions of attempts after disclosure
If your website runs on PHP, and there's a good chance it does, update it now. Security researchers have disclosed PHP2Shell, a set of critical flaws in the core of PHP, the programming language that runs roughly 70% of websites, including WordPress, Drupal, Joomla and Wikipedia's MediaWiki.
PHP released a patch on October 21, 2026. The bugs are tracked as CVE-2026-109618 and CVE-2026-109619, with severity scores up to 9.2 out of 10.
What PHP2Shell can do
According to Wiz Research, which found the flaws, attackers can:
- run their own code on a server without logging in (remote code execution), and
- read the server's memory, which may contain passwords, login tokens, API keys and certificates.
The problem sits in PHP itself, not in any one website's code. Wiz says even a simple "Hello World" page on an affected PHP version can be vulnerable. Every PHP release from the past decade, back to version 5.4, is hit by at least one of the flaws.
Where AI comes in
The bugs weren't new. Wiz says the underlying flaws had been in the code for more than a decade. They come from how PHP cleans up after a request fails, especially when it runs out of memory, which can leave broken internal state for the next request to trip over.
What changed is the effort needed to exploit them. Building a working attack meant chaining several bugs and adjusting to each server's memory layout through endless testing. Wiz says that used to sit "at the edge of what human researchers could practically achieve." With frontier AI models, it built and adapted working exploits in hours. The team says it used Gemini 4 Argon and GPT-5.6 Sol, along with its own tools.
"We believe PHP2Shell could mark a similar turning point for the AI era." Wiz Research, comparing it with Heartbleed in 2014 and Log4Shell in 2021
ADVERTISEMENT
Attacks are already happening
This isn't theoretical. Reporting by The Hacker News, citing several security firms, describes:
- Millions of exploitation attempts detected by Cloudflare after public disclosure.
- A European government entity breached; Palo Alto Networks' Unit 42 says attackers reached password hashes and two-factor secrets within two hours, and their commands looked like the work of an AI agent.
- A suspected North Korea-linked campaign, according to CYFIRMA, quietly scanning 216 unpatched hosts across government, finance, aviation and other sectors.
Amazon has also said its honeypots recorded attempts on October 17, four days before the flaws were made public, and that it's unclear how the attacker got the information.
What you should do
- Update PHP to the patched release for your version, including older branches you may still run.
- Ask your host whether they've patched, if you use shared hosting or a managed WordPress service.
- Rotate secrets (passwords, API keys) if your site was exposed and unpatched.
Major providers, including Cloudflare, AWS, Google, Microsoft, Akamai and Fastly, prepared temporary protections for sites that can't patch immediately, but a filter is no substitute for the update.
Why it matters
AI is cutting the time between "this bug exists" and "this bug is being used against you." It's the same race Anthropic described when it launched its Cyber Mission, and the reason researchers are now using AI to prove code is safe instead of just testing it.
Sources
ScoviAI Team is the ScoviAI newsroom. We write every article ourselves from the original announcements and link our sources. Spotted a mistake? Tell us and we'll fix it.
ADVERTISEMENT
