DEV Community

Abdullah Ahmad
Abdullah Ahmad

Posted on

AI just uncovered critical security flaws that stayed hidden in PHP for over a decade.

Just few days ago, on October 21, researchers disclosed two critical vulnerabilities in PHP, a programming language used by millions of websites, including sites running WordPress, Drupal, and Joomla.

This wasn't just a routine patch, it triggered an unprecedented global scramble, putting over 70% of the web at risk and forcing government portals worldwide to go completely dark to prevent serious attacks and damages.

The surprising part? The underlying weaknesses had existed for more than a decade, 2012.

Both vulnerabilities involved how PHP handled memory under certain error conditions. In short, the software could end up using memory that was no longer valid. With carefully crafted requests, an attacker could potentially execute code on a vulnerable server without logging in, or access sensitive information from its memory.

But how did AI help uncover them?

Researchers at Wiz used AI models to investigate PHP's code, test possible weaknesses, and help develop and debug complex exploits.

Calif, working with OpenAI, also identified one of the critical flaws using GPT-5.6 Sol.

The difficult part wasn't simply finding suspicious code.
It was understanding how several low-level memory issues could interact and turn into a working attack. AI helped researchers work through that complexity.

PHP maintainers released security fixes on October 21, following coordinated work with researchers and security teams.

What makes this incident important is that AI helped uncover and demonstrate vulnerabilities that had remained difficult to exploit for years.

Cybersecurity #ArtificialIntelligence #Infosec #PHP #TechNews

Top comments (0)