OpenAI and Google Uncover Critical Vulnerability in PHP

Key Takeaways

1OpenAI and Google models have identified a critical vulnerability in PHP

2A patch has been coordinated over two months with developers, Wiz, Cloudflare, and U.S. agencies

3Wiz emphasizes that LLMs make it easier to adapt exploits to different configurations

4A GitHub tool allows users to check if a PHP installation is affected

Why it matters : This discovery illustrates both the potential of AI for security and the risk that it also facilitates attacks.

OpenAI and Google models have uncovered a critical vulnerability in PHP that could affect millions of websites. The response has mobilized developers, companies, and U.S. agencies for two months, while security experts emphasize that the same AI models can also accelerate the development of exploits. A tool on GitHub already allows users to check for exposed installations.

OpenAI anticipates more "PHP2Shell" cases and warns about offensive use

OpenAI expects a significant increase in discoveries akin to "PHP2Shell" in the coming months. The company warns that the same AI capabilities that identify vulnerabilities can also be used by attackers. This situation fuels an ongoing debate about the offensive and defensive uses of AI models in cybersecurity. OpenAI considers this vulnerability one of the most significant security issues in recent years.

A five-phase patch coordinated with CISA and Cloudflare

The patch was coordinated over two months, in five stages, with the participation of PHP developers, Wiz, Cloudflare, and U.S. agencies such as CISA and the White House. OpenAI suggests using this process as a model for future AI-assisted security discoveries.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

Joint discovery and GitHub tool to check versions

OpenAI and Google models have identified a critical vulnerability in PHP that could affect millions of websites worldwide. GPT-5.6 Sol highlighted this flaw within the Patch-the-Planet initiative. A utility hosted on GitHub provides server administrators with the ability to check if their PHP version is vulnerable. OpenAI indicates that some vulnerabilities were spotted before attackers could exploit them.

According to Wiz, LLMs lower the adaptation threshold for exploits

The security company Wiz independently discovered the same vulnerabilities using its Wiz Atlas platform, combined with Gemini 4 Argon and GPT-5.6 Sol. Wiz specifies that Gemini 4 Argon proved particularly effective in spotting the vulnerability, generating exploits, developing attack detection methods, and confirming the validity of the patch. According to Wiz, adapting a PHP exploit to other versions and configurations previously required significant effort. Now, large language models can automatically adjust these exploits according to versions, configurations, and memory layouts, significantly reducing the difficulty of extending attacks and making previously impractical vulnerabilities exploitable.