OpenAI has open-sourced Codex Security CLI, a command-line tool designed to find and fix vulnerabilities in code repositories.
The Decoder reports that the system was previously known internally as Aardvark and has already helped fix more than 3,000 critical security flaws, according to OpenAI. The tool arrives as AI companies race to apply coding models to defensive security work.
A command-line interface matters because it can fit into the way developers already inspect, test, and patch code. If the tool can produce useful fixes rather than only flagging problems, it could reduce the gap between vulnerability discovery and remediation.
The limits are important. Automated fixes still need review, tests, and careful handling of false positives. But the release shows how frontier AI companies are turning security agents into developer products, not just internal research demonstrations.