# OpenAI Codex Security Scans 1.2 Million Commits, Flags Over 10,000 High-Severity Vulnerabilities

*2026-03-09 23:01 UTC · Technology · 1 min read*

> OpenAI launched Codex Security, an AI agent that scanned 1.2 million commits in its first month, uncovering over 10,000 high-severity bugs including 14 new CVEs.

OpenAI launched Codex Security, an AI-powered security agent that autonomously finds, validates, and proposes fixes for software vulnerabilities. During its first 30 days of research testing, the tool scanned over 1.2 million commits across external repositories, flagging 10,561 high-severity issues and 792 critical findings.

Among the vulnerabilities discovered were flaws in widely used open-source projects including OpenSSH, GnuTLS, PHP, Chromium, and libssh, with 14 CVEs assigned so far. The tool evolved from Aardvark, OpenAI's private beta vulnerability scanner from October 2025, and has achieved a 50% reduction in false positive rates over its testing period.

Codex Security is now available in research preview to ChatGPT Pro, Enterprise, Business, and Edu customers, with free usage for the first month. The launch positions OpenAI deeper into enterprise security workflows, competing with established players in the application security market.

## Sources

- [thehackernews.com](https://thehackernews.com/2026/03/openai-codex-security-scanned-12.html)
- [openai.com](https://openai.com/index/codex-security-now-in-research-preview/)
- [csoonline.com](https://www.csoonline.com/article/4142354/openai-says-codex-security-found-11000-high-impact-bugs-in-a-month.html)

---
TechTools News · https://techtools.cz/tools/news/article/683-openai-codex-security-scans-1-2-million-commits-flags-over-10-000-high-severity
Česká verze: https://techtools.cz/tools/news/clanek/696-codex-security-od-openai-prohledal-1-2-milionu-commitu-a-odhalil-pres-10-000
