OpenAI Codex Security Uncovers 10,561 High-Severity Issues in 1.2 Million Commits

OpenAI rolled out Codex Security in a research preview for ChatGPT Pro/Enterprise customers, offering AI-driven code security checks that analyze a project’s context, generate a threat model, validate vulnerabilities in a sandbox, and propose fixes. In beta, it scanned 1.2 million commits across external repos and found 792 critical and 10,561 high-severity issues, including flaws in OpenSSH, GnuTLS, GOGS, Thorium, libssh, PHP, and Chromium. OpenAI says the system reduces false positives through context grounding and automated validation, with free usage for the next month.