OpenAI Discovers Vulnerability in PHP Using AI Model
October 24, 2026 · 1 min read
OpenAI announced on October 21 that it has discovered a vulnerability in the open-source programming language PHP using its AI model. A revised version of PHP was released on the same day. The organization predicts that similar discoveries may increase significantly over the coming months.
The discovery is part of OpenAI's security support project called "Patch the Planet," where the security company Calif utilized the AI model "GPT-5.6 Sol" for this finding. OpenAI collaborated with partner companies and the U.S. government over a two-month period to prepare for the release of the PHP vulnerability and its fix.
Additionally, the security company Wiz independently discovered this vulnerability. According to Wiz, they employed not only GPT-5.6 Sol but also Google's AI model "Gemini 4 Argon."
OpenAI pointed out that while this vulnerability was identified by cutting-edge AI before attackers could exploit it, attackers can also misuse AI. OpenAI is addressing other severe vulnerabilities and expects similar discoveries to rise in the coming months.
In light of this, OpenAI is working on providing support through its security program and building frameworks for vulnerability mitigation based on AI.