OpenAI Open-Sources Codex Security: AI Scanner Finds 792 Critical Vulns in 1.2M Commits
OpenAI released Codex Security as an open-source CLI and TypeScript SDK under Apache-2.0. The tool uses AI to automatically scan code repositories for vulnerabilities, verify them, and generate patches. Previously a closed-source Codex plugin, it was made independent after community requests. The standalone version supports batch scanning across organizations, history, deduplication, false-positive tracking, and CI integration. It defaults to the gpt-5.6-sol model and requires Node.js 22+ and Python 3.10+. In public testing across 1.2 million commits, it found 792 critical and 10,561 high-severity vulnerabilities in projects like OpenSSH, GnuTLS, PHP, and Chromium. A third-party comparison on 162,000 lines of production code showed a 74% true positive rate, versus 28% for Snyk and 20% for Semgrep.