OpenAI says PHP flaws found with GPT-5.6 Sol are major vulnerabilities and similar findings are set to surge.
gpt-5 openai
| Source: Techmeme | Original article
OpenAI reports that using its GPT‑5.6 Sol tool it uncovered significant PHP vulnerabilities and expects the number of similar discoveries to rise sharply.
OpenAI confirmed that its GPT‑5.6 Sol model helped uncover one of two critical PHP vulnerabilities that sparked an emergency patch, worldwide government alerts and temporary shutdowns of numerous websites. In a blog post titled “Patch the Planet: global response coordination as agents uncover critical vulnerabilities,” the company said the flaws flagged by GPT‑5.6 Sol are “among several significant vulnerabilities” it has identified and warned that “similar findings to increase sharply.”
The discovery was made by Calif, a cybersecurity research firm that used GPT‑5.6 Sol as part of OpenAI’s Patch the Planet initiative, a two‑month effort that brought together maintainers, researchers and security providers to remediate the weaknesses. A second firm, Wiz, independently reported both PHP bugs, confirming their severity. The vulnerabilities had lingered in the language’s core code for years, exposing web applications to remote code execution and data leakage. Their exposure prompted coordinated patches across the PHP ecosystem and a wave of advisories from national cyber‑security agencies.
The episode underscores a growing shift: AI models are moving from code‑completion tools to active participants in vulnerability research. OpenAI’s statement suggests it now views large‑scale AI‑driven scanning as a routine part of its security workflow, a stance echoed by recent reports of AI agents flagging hundreds of issues in open‑source projects.
Going forward, the industry will watch how OpenAI scales the Patch the Planet model and whether other AI providers follow suit. Regulators may also scrutinise the balance between rapid disclosure and responsible coordination, especially as AI‑identified flaws could surface at unprecedented speed. Stakeholders are likely to see an uptick in AI‑assisted security tooling, heightened collaboration among vendors, and a push for standards governing AI‑driven vulnerability discovery.
Sources
Back to AIPULSEN