OpenAI Codex Security Scans 1.2M Commits, Uncovers 10K Critical Bugs

OpenAI has rolled out Codex Security, an AI‑driven assistant that automatically scans code repositories for complex vulnerabilities, validates the findings, and suggests remediation steps. As part of a free one‑month preview for eligible ChatGPT customers, the tool examined roughly 1.2 million commits across diverse projects and identified 10,561 critical flaws, many of which were previously undiscovered.

For defenders, the results underscore how AI can amplify vulnerability discovery at scale, revealing a hidden attack surface that traditional static analysis often misses. Integrating similar AI‑powered scanning into your CI/CD pipeline can accelerate detection, prioritize remediation, and reduce exposure before adversaries exploit these weaknesses. Stay aware of the preview’s limitations—false positives and coverage gaps—and plan to incorporate robust verification and patch management processes to maximize the benefit.

Categories: AI Security & Threats, Vulnerabilities & Exploits

Source: Read original article